emnlp emnlp2012 emnlp2012-118 knowledge-graph by maker-knowledge-mining

118 emnlp-2012-Source Language Adaptation for Resource-Poor Machine Translation

Source: pdf

Author: Pidong Wang ; Preslav Nakov ; Hwee Tou Ng

Abstract: We propose a novel, language-independent approach for improving machine translation from a resource-poor language to X by adapting a large bi-text for a related resource-rich language and X (the same target language). We assume a small bi-text for the resourcepoor language to X pair, which we use to learn word-level and phrase-level paraphrases and cross-lingual morphological variants between the resource-rich and the resource-poor language; we then adapt the former to get closer to the latter. Our experiments for Indonesian/Malay–English translation show that using the large adapted resource-rich bitext yields 6.7 BLEU points of improvement over the unadapted one and 2.6 BLEU points over the original small bi-text. Moreover, combining the small bi-text with the adapted bi-text outperforms the corresponding combinations with the unadapted bi-text by 1.5– 3 BLEU points. We also demonstrate applicability to other languages and domains.

Reference: text

Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 s g Abstract We propose a novel, language-independent approach for improving machine translation from a resource-poor language to X by adapting a large bi-text for a related resource-rich language and X (the same target language). [sent-4, score-0.088]

2 We assume a small bi-text for the resourcepoor language to X pair, which we use to learn word-level and phrase-level paraphrases and cross-lingual morphological variants between the resource-rich and the resource-poor language; we then adapt the former to get closer to the latter. [sent-5, score-0.141]

3 Our experiments for Indonesian/Malay–English translation show that using the large adapted resource-rich bitext yields 6. [sent-6, score-0.115]

4 7 BLEU points of improvement over the unadapted one and 2. [sent-7, score-0.047]

5 Moreover, combining the small bi-text with the adapted bi-text outperforms the corresponding combinations with the unadapted bi-text by 1. [sent-9, score-0.091]

6 1 Introduction Statistical machine translation (SMT) systems learn how to translate from large sentence-aligned bilingual corpora ofhuman-generated translations, called bi-texts. [sent-12, score-0.053]

7 , X=English) to improve machine translation from a resource-poor language to X (Nakov and Ng, 2009; Nakov and Ng, 2012). [sent-27, score-0.053]

8 We assume a small bi-text for the resource-poor language, which we use to learn word-level and phrase-level paraphrases and cross-lingual morpho- logical variants between the two languages. [sent-29, score-0.051]

9 Assuming translation into the same target language X, we adapt (the source side of) a large training bi-text for a related resource-rich language and X. [sent-30, score-0.074]

10 Training on the adapted large bi-text yields very significant improvements in translation quality compared to both (a) training on the unadapted version, and (b) training on the small bi-text for the resourcepoor language. [sent-31, score-0.164]

11 We further achieve very sizable improvements when combining the small bi-text with the large adapted bi-text, compared to combining the former with the unadapted bi-text. [sent-32, score-0.091]

12 In contrast, we have a different objective we do not carry out full translation but rather adaptation since our ultimate goal is to translate into a third language X. [sent-39, score-0.08]

13 A special case of this same line of research is the translation between dialects of the same language, e. [sent-40, score-0.087]

14 , between Cantonese and Mandarin (Zhang, 1998), or between a dialect of a language and a standard version of that language, e. [sent-42, score-0.04]

15 (201 1) described a rule-based system for adapting Brazilian Portuguese (BP) to European Portuguese (EP), which they used to adapt BP–English bi-texts to EP–English. [sent-59, score-0.056]

16 They report small improvements in BLEU for EP–English translation when training on the adapted “EP”–En bi-text compared to using the unadapted BP–En (38. [sent-60, score-0.144]

17 Another alternative, which we do not explore in this work, is to use cascaded translation using a pivot language (Utiyama and Isahara, 2007; Cohn and Lapata, 2007; Wu and Wang, 2009). [sent-72, score-0.148]

18 Unfortunately, using the resource-rich language as a pivot (poor→rich→X) would require an additional parall(pelo poor–rich bi-text, wulhdic rheq we deo a nno atd hdaitvioen. [sent-73, score-0.095]

19 Consider, for example, Article 1 of the Universal Declaration of Human Rights:2 • Semua manusia dilahirkan bebas dan samarata dari segi kemuliaan dan hak-hak. [sent-95, score-0.135]

20 Mereka mempunyai pemikiran dan perasaan hati dan hendaklah bertindak di antara satu sama lain dengan semangat persaudaraan. [sent-96, score-0.246]

21 (Malay) • Semua orang dilahirkan merdeka dan mempunyai martabat dan hak-hak yang sama. [sent-97, score-0.103]

22 Mereka dikaruniai akal dan hati nurani dan hendaknya bergaul satu sama lain dalam semangat persaudaraan. [sent-98, score-0.207]

23 Other differences are due to the use of different morphological forms, e. [sent-102, score-0.049]

24 Of course, word choice in translation is often a matter of taste. [sent-106, score-0.053]

25 Thus, we asked a native speaker of Indonesian to adapt the Malay version to Indonesian while preserving as many words as possible: • Semua manusia dilahirkan bebas dan mempunyai martabat dan hak-hak yang sama. [sent-107, score-0.156]

26 Mereka mempunyai pemikiran dan perasaan dan hendaklah bergaul satu sama lain dalam semangat persaudaraan. [sent-108, score-0.262]

27 4 Method We improve machine translation from a resourcepoor language (Indonesian) to English by adapting a bi-text for a related resource-rich language (Malay) and English, using word-level and phrase-level paraphrases and cross-lingual morphological variants. [sent-118, score-0.182]

28 1 Word-Level Paraphrasing Given a Malay sentence, we generate a confusion network containing multiple Indonesian word-level paraphrase options for each Malay word. [sent-120, score-0.153]

29 Each such Indonesian option is associated with a corresponding weight in the network, which is defined as the probability of this option being a translation of the original Malay word (see Eq. [sent-121, score-0.053]

30 We decode this confusion network using a large Indonesian language model, thus generating a ranked list of n corresponding adapted “Indonesian” sentences. [sent-123, score-0.151]

31 Then, we pair each such adapted “Indonesian” sentence with the English counter-part for the Malay sentence it was derived from, thus obtaining a synthetic “Indonesian”–English bi-text. [sent-124, score-0.06]

32 Below we first describe how we generate wordlevel Indonesian options and corresponding weights for the Malay words. [sent-126, score-0.046]

33 Then, we explain how we build, decode, and improve the confusion network. [sent-127, score-0.069]

34 , a correct translation for the Malay samarata can be obtained by splitting it into the Indonesian sequence sama rata. [sent-130, score-0.123]

35 1 Inducing Word-Level Paraphrases We use pivoting over English to induce potential Indonesian translations for a given Malay word. [sent-134, score-0.054]

36 Then, we induce Indonesian-Malay word translation pairs assuming that if an Indonesian word i and a Malay word m are aligned to the same English word e, they could be mutual translations. [sent-136, score-0.053]

38 2 Confusion Network Construction Given a Malay sentence, we construct an Indonesian confusion network, where each Malay word is augmented with a set of network transitions: possible Indonesian word translations. [sent-142, score-0.107]

39 The weight of such a transition is the conditional IndonesianMalay translation probability as calculated by Eq. [sent-143, score-0.053]

40 This is necessary because of the large number of false friends and partial cognates between Malay and Indonesian (see Section 3). [sent-148, score-0.058]

41 Finally, we decode the confusion network for a Malay sentence using a large Indonesian language model, and we extract an n-best list. [sent-149, score-0.107]

42 5 Table 1 shows the 10-best adapted “Indonesian” sentences6 we generated for the confusion network in Figure 1. [sent-150, score-0.151]

43 6According to a native Indonesian speaker, options 1 and 3 in Table 1are perfect adaptations, options 2 and 5 have a wrong word order, and the rest are grammatical though not perfect. [sent-155, score-0.092]

44 289 Moreover, the options we propose for a Malay word are inherently restricted to the small Indonesian vocabulary of the Indonesian–English bi-text. [sent-156, score-0.066]

45 We increase the Indonesian options for a Malay word using morphology. [sent-172, score-0.046]

46 Since the set of Indonesian options for a Malay word in pivoting is restricted to the Indonesian vocabulary of the small Indonesian– English bi-text, this is a severe limitation of pivoting. [sent-173, score-0.12]

47 Since Malay and Indonesian are subject to the same morphological processes and share many lemmata, we use such groups to propose Indonesian translation options for a Malay word. [sent-180, score-0.148]

48 The union of these groups is the set of morphological variants that we will add to the confusion network as additional options for the Malay word. [sent-184, score-0.228]

49 2 Phrase-Level Paraphrasing Word-level paraphrasing ignores context when generating Indonesian variants, relying on the Indonesian language model to make the right contextual choice. [sent-188, score-0.072]

50 We also try to model context more directly by generating adaptation options at the phrase level. [sent-189, score-0.112]

51 7While the different morphological forms typically have different meanings, e. [sent-190, score-0.049]

52 peminum (‘drinker’), in some cases the forms could have the same translation in English, e. [sent-193, score-0.069]

53 This is our motivation for trying morphological variants, even though they are almost exclusively derivational, and thus quite risky as translational variants; see also (Nakov and Ng, 2011). [sent-197, score-0.049]

54 We use standard phrase-based SMT techniques to build separate phrase tables for the Indonesian–English and the Malay–English bi-texts, where we have four conditional probabilities: forward/reverse phrase translation probability, and forward/reverse lexicalized phrase translation probability. [sent-199, score-0.191]

55 We pivot over English to generate Indonesian-Malay phrase pairs, whose probabilities are derived from the corresponding ones in the two phrase tables using Eq. [sent-200, score-0.158]

56 While phrase-level paraphrasing models context better, it remains limited in the size of its Indonesian vocabulary by the small Indonesian–English bi-text, just like word-level paraphrasing was. [sent-203, score-0.164]

57 We address this by transforming the sentences in the development and the test Indonesian–English bi-texts into confusion networks, where we add Malay morphological variants for the Indonesian words, weighting them as before. [sent-204, score-0.144]

58 Moreover, the original Malay–English bi-text is much larger in size than the Indonesian–English one, and now it has been further expanded n times in order to become an “Indonesian”–English bi-text, which means that it will dominate the concatenation due to its size. [sent-214, score-0.043]

59 Finally, we experiment with a method for combining phrase tables proposed in (Nakov and Ng, 2009; Nakov and Ng, 2012). [sent-217, score-0.041]

60 The first phrase table is extracted from word alignments for the balanced concatenation with repetitions, which are then truncated so that they are kept for only one copy of the Indonesian–English bi-text. [sent-218, score-0.113]

61 The two tables are then merged as follows: all phrase pairs from the first one are retained, and to them are added those phrase pairs from the second one that are not present in the first one. [sent-220, score-0.063]

62 Other possibilities for combining the phrase tables include using alternative decoding paths (Birch et al. [sent-231, score-0.041]

63 2 Baseline Systems We build five baseline gle bi-text, ML2EN or ing ML2EN and IN2EN, balanced concatenation, systems two using a sinIN2EN, and three combinusing simple concatenation, and sophisticated phrase ta– ble combination. [sent-243, score-0.049]

64 3 Isolated Experiments The isolated experiments only use the adapted “Indonesian”–English bi-text, which allows for a direct comparison to using ML2EN / IN2EN only. [sent-246, score-0.08]

65 1 Word-Level Paraphrasing In our word-level paraphrasing experiments, we adapt Malay to Indonesian using three kinds of confusion networks (see Section 4. [sent-249, score-0.186]

66 3 for details): • CN:pivot using word-level pivoting only; • CN:pivot′ using owrdo-rldev-leevle pil pivoting, w;ith probabilities from word alignments for IN2EN that were improved using ML2EN; • CN:pivot′+morph CN:pivot′ augmented with cross-lingual morphological variants. [sent-251, score-0.124]

67 2 Phrase-Level Paraphrasing In our phrase-level paraphrasing experiments, we use pivoted phrase tables (PPT) with the following features for each phrase table entry (in addition to the phrase penalty; see Section 4. [sent-261, score-0.157]

68 We experiment with the same three combinations: simple concatenation, balanced concatenation, and sophisticated phrase table combina- tion. [sent-266, score-0.049]

69 While simple concatenation does not help, balanced concatenation with repetitions improves by 1. [sent-283, score-0.129]

70 We can see that word-level paraphrasing improves by up to 5. [sent-289, score-0.072]

71 When morphological variants are further added, the unigram precision improves by almost 1% absolute over CN:pivot′. [sent-314, score-0.075]

72 The lower part of Table 3 shows that phrase-level paraphrasing performs a bit better. [sent-316, score-0.072]

73 Extending the Indonesian vocabulary with cross-lingual morphological variants is still helpful, though not as much as at the word-level. [sent-319, score-0.095]

74 3 for additional details) when applied to combine IN2EN (1) with the original ML2EN and (2) with various adapted versions of it. [sent-326, score-0.044]

75 We can see that for the word-level paraphrasing experiments (CN: *), all combinations except for CN:pivot perform significantly better than their corresponding baselines, but the improvements are most sizeable for the simple concatenation. [sent-327, score-0.088]

76 31 BLEU points between the balanced concatenation and the sophisticated combination for the original ML2EN, they differ little for the adapted versions. [sent-329, score-0.114]

77 This is probably due to the sophisticated combination as- suming that the second bi-text is worse than the first one, which is not really the case for the adapted versions: as Table 3 shows, they all outperform IN2EN. [sent-330, score-0.044]

78 Overall, phrase-level paraphrasing performs a bit better than word-level paraphrasing, and system combination with MEMT improves even further. [sent-331, score-0.072]

79 In CN: * above, we paraphrased each word in the Malay input, because of false friends like polisi and partial cognates like nanti. [sent-334, score-0.074]

80 , changing beliau (‘he’, respectful) to ia (‘he’, casual), which confusion network weights and LM would not always handle. [sent-337, score-0.107]

81 Since IN-LM occasionally contains some Malay-specific words, we also tried paraphrasing words that occur at most t times in IN-LM. [sent-341, score-0.072]

82 , to adapt Indonesian to Malay: we thus build a “Malay” confusion network for each dev/test Indonesian sentence to be used as an input to a Malay–English SMT system trained on the ML2EN dataset. [sent-365, score-0.128]

83 We tried two variations of this idea: • lattice: Use Indonesian-to-Malay confusion nlaetttwicoer:ks directly as input to the ML2EN SMT system, i. [sent-366, score-0.069]

84 , tune a log-linear model using con- • fusion networks for the source side of the IN2EN-dev dataset, and then evaluate the tuned system using confusion networks for the source side of the IN2EN-test dataset. [sent-368, score-0.133]

85 1-best: Use the 1-best output from the Indonesian-to-Malay confusion network for each sentence of IN2EN-dev and IN2EN-test. [sent-369, score-0.107]

86 Finally, get an adapted “Malay”–English development set and an adapted “Malay”–English test set, and use them to tune and evaluate the ML2EN SMT system. [sent-371, score-0.104]

87 In contrast, CN:pivot uses both n-best outputs and an Indonesian LM; designing a similar setup for reversed adaptation is a research direction we would like to pursue in future work. [sent-374, score-0.045]

88 Table 8 shows that both CN: * and PPT: * yield statistically significant improvements over balanced concatenation with unadapted BG2EN; system combination with MEMT improves even further. [sent-387, score-0.117]

89 2961 56 Balanced concatenation of MK2EN with an adapted BG2EN Table 8: Improving Macedonian–English adapting Bulgarian to Macedonian. [sent-401, score-0.122]

90 295 8 Conclusion and Future Work We have presented a novel approach for improving machine translation for a resource-poor language by adapting a bi-text for a related resource-rich language, using confusion networks, word/phrase-level paraphrasing, and morphological analysis. [sent-405, score-0.206]

91 7 BLEU points over an unadapted version of ML2EN, 2. [sent-407, score-0.047]

92 In future work, we would like to add word deletion, insertion, splitting, and concatenation as allowed editing operations. [sent-411, score-0.043]

93 A machine translation system between a pair of closely related languages. [sent-419, score-0.053]

94 Machine translation by triangulation: Making effective use of multiparallel corpora. [sent-443, score-0.053]

95 Combining machine translation output with open source: The Carnegie Mellon multi-engine machine translation scheme. [sent-459, score-0.106]

96 Improved statistical machine translation for resource-poor languages using related resource-rich languages. [sent-467, score-0.071]

97 Improving statistical machine translation for a resource-poor language using related resource-rich languages. [sent-475, score-0.053]

98 Combining word-level and character-level models for machine translation between closely-related languages. [sent-479, score-0.053]

99 Dialectal to Standard Arabic paraphrasing to improve Arabic- English statistical machine translation. [sent-491, score-0.072]

100 A comparison of pivot methods for phrase-based statistical machine translation. [sent-509, score-0.095]

similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('indonesian', 0.737), ('malay', 0.572), ('pivot', 0.095), ('nakov', 0.085), ('cn', 0.08), ('paraphrasing', 0.072), ('macedonian', 0.071), ('confusion', 0.069), ('pivoting', 0.054), ('translation', 0.053), ('bleu', 0.052), ('smt', 0.051), ('morphological', 0.049), ('unadapted', 0.047), ('ppt', 0.047), ('options', 0.046), ('adapted', 0.044), ('concatenation', 0.043), ('dialect', 0.04), ('minum', 0.039), ('sama', 0.039), ('english', 0.039), ('network', 0.038), ('cognates', 0.037), ('isolated', 0.036), ('arabic', 0.035), ('adapting', 0.035), ('dialects', 0.034), ('preslav', 0.034), ('bulgarian', 0.032), ('drink', 0.031), ('mempunyai', 0.031), ('minuman', 0.031), ('morph', 0.031), ('samarata', 0.031), ('balanced', 0.027), ('adaptation', 0.027), ('egyptian', 0.027), ('variants', 0.026), ('paraphrases', 0.025), ('networks', 0.024), ('dilahirkan', 0.024), ('hendaklah', 0.024), ('lain', 0.024), ('memt', 0.024), ('mereka', 0.024), ('satu', 0.024), ('semangat', 0.024), ('semua', 0.024), ('phrase', 0.022), ('alignments', 0.021), ('friends', 0.021), ('portuguese', 0.021), ('adapt', 0.021), ('adaptations', 0.02), ('resourcepoor', 0.02), ('ep', 0.02), ('vocabulary', 0.02), ('singapore', 0.02), ('tables', 0.019), ('reversed', 0.018), ('mk', 0.018), ('bitext', 0.018), ('ng', 0.018), ('languages', 0.018), ('pr', 0.017), ('try', 0.017), ('lm', 0.016), ('tune', 0.016), ('synthetic', 0.016), ('dan', 0.016), ('repetitions', 0.016), ('altintas', 0.016), ('bakr', 0.016), ('bebas', 0.016), ('bergaul', 0.016), ('brazilian', 0.016), ('conscience', 0.016), ('dalam', 0.016), ('eentnegnlicesh', 0.016), ('gaelic', 0.016), ('hati', 0.016), ('hendaknya', 0.016), ('irish', 0.016), ('lemmata', 0.016), ('malaysia', 0.016), ('manusia', 0.016), ('martabat', 0.016), ('marujo', 0.016), ('opus', 0.016), ('pada', 0.016), ('pemikiran', 0.016), ('peminum', 0.016), ('perasaan', 0.016), ('polisi', 0.016), ('salloum', 0.016), ('segi', 0.016), ('sizeable', 0.016), ('tahun', 0.016), ('hwee', 0.016)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 1.0 118 emnlp-2012-Source Language Adaptation for Resource-Poor Machine Translation

Author: Pidong Wang ; Preslav Nakov ; Hwee Tou Ng

2 0.050983433 58 emnlp-2012-Generalizing Sub-sentential Paraphrase Acquisition across Original Signal Type of Text Pairs

Author: Aurelien Max ; Houda Bouamor ; Anne Vilnat

Abstract: This paper describes a study on the impact of the original signal (text, speech, visual scene, event) of a text pair on the task of both manual and automatic sub-sentential paraphrase acquisition. A corpus of 2,500 annotated sentences in English and French is described, and performance on this corpus is reported for an efficient system combination exploiting a large set of features for paraphrase recognition. A detailed quantified typology of subsentential paraphrases found in our corpus types is given.

3 0.041926105 86 emnlp-2012-Locally Training the Log-Linear Model for SMT

Author: Lemao Liu ; Hailong Cao ; Taro Watanabe ; Tiejun Zhao ; Mo Yu ; Conghui Zhu

Abstract: In statistical machine translation, minimum error rate training (MERT) is a standard method for tuning a single weight with regard to a given development data. However, due to the diversity and uneven distribution of source sentences, there are two problems suffered by this method. First, its performance is highly dependent on the choice of a development set, which may lead to an unstable performance for testing. Second, translations become inconsistent at the sentence level since tuning is performed globally on a document level. In this paper, we propose a novel local training method to address these two problems. Unlike a global training method, such as MERT, in which a single weight is learned and used for all the input sentences, we perform training and testing in one step by learning a sentencewise weight for each input sentence. We pro- pose efficient incremental training methods to put the local training into practice. In NIST Chinese-to-English translation tasks, our local training method significantly outperforms MERT with the maximal improvements up to 2.0 BLEU points, meanwhile its efficiency is comparable to that of the global method.

4 0.038317196 128 emnlp-2012-Translation Model Based Cross-Lingual Language Model Adaptation: from Word Models to Phrase Models

Author: Shixiang Lu ; Wei Wei ; Xiaoyin Fu ; Bo Xu

Abstract: In this paper, we propose a novel translation model (TM) based cross-lingual data selection model for language model (LM) adaptation in statistical machine translation (SMT), from word models to phrase models. Given a source sentence in the translation task, this model directly estimates the probability that a sentence in the target LM training corpus is similar. Compared with the traditional approaches which utilize the first pass translation hypotheses, cross-lingual data selection model avoids the problem of noisy proliferation. Furthermore, phrase TM based cross-lingual data selection model is more effective than the traditional approaches based on bag-ofwords models and word-based TM, because it captures contextual information in modeling the selection of phrase as a whole. Experiments conducted on large-scale data sets demonstrate that our approach significantly outperforms the state-of-the-art approaches on both LM perplexity and SMT performance.

5 0.037597965 39 emnlp-2012-Enlarging Paraphrase Collections through Generalization and Instantiation

Author: Atsushi Fujita ; Pierre Isabelle ; Roland Kuhn

Abstract: This paper presents a paraphrase acquisition method that uncovers and exploits generalities underlying paraphrases: paraphrase patterns are first induced and then used to collect novel instances. Unlike existing methods, ours uses both bilingual parallel and monolingual corpora. While the former are regarded as a source of high-quality seed paraphrases, the latter are searched for paraphrases that match patterns learned from the seed paraphrases. We show how one can use monolingual corpora, which are far more numerous and larger than bilingual corpora, to obtain paraphrases that rival in quality those derived directly from bilingual corpora. In our experiments, the number of paraphrase pairs obtained in this way from monolingual corpora was a large multiple of the number of seed paraphrases. Human evaluation through a paraphrase substitution test demonstrated that the newly acquired paraphrase pairs are ofreasonable quality. Remaining noise can be further reduced by filtering seed paraphrases.

6 0.037095226 35 emnlp-2012-Document-Wide Decoding for Phrase-Based Statistical Machine Translation

7 0.036181279 54 emnlp-2012-Forced Derivation Tree based Model Training to Statistical Machine Translation

8 0.03544319 42 emnlp-2012-Entropy-based Pruning for Phrase-based Machine Translation

9 0.031398989 11 emnlp-2012-A Systematic Comparison of Phrase Table Pruning Techniques

10 0.030854424 135 emnlp-2012-Using Discourse Information for Paraphrase Extraction

11 0.029742356 1 emnlp-2012-A Bayesian Model for Learning SCFGs with Discontiguous Rules

12 0.028711582 127 emnlp-2012-Transforming Trees to Improve Syntactic Convergence

13 0.027380131 31 emnlp-2012-Cross-Lingual Language Modeling with Syntactic Reordering for Low-Resource Speech Recognition

14 0.025936555 67 emnlp-2012-Inducing a Discriminative Parser to Optimize Machine Translation Reordering

15 0.025013085 108 emnlp-2012-Probabilistic Finite State Machines for Regression-based MT Evaluation

16 0.023889625 16 emnlp-2012-Aligning Predicates across Monolingual Comparable Texts using Graph-based Clustering

17 0.022954367 123 emnlp-2012-Syntactic Transfer Using a Bilingual Lexicon

18 0.022774426 109 emnlp-2012-Re-training Monolingual Parser Bilingually for Syntactic SMT

19 0.021240268 32 emnlp-2012-Detecting Subgroups in Online Discussions by Modeling Positive and Negative Relations among Participants

20 0.020885337 79 emnlp-2012-Learning Syntactic Categories Using Paradigmatic Representations of Word Context

similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, 0.081), (1, -0.046), (2, -0.088), (3, 0.005), (4, -0.006), (5, 0.008), (6, 0.004), (7, -0.005), (8, 0.012), (9, -0.039), (10, 0.019), (11, -0.005), (12, -0.005), (13, -0.002), (14, -0.008), (15, 0.005), (16, 0.012), (17, 0.01), (18, -0.001), (19, 0.047), (20, -0.046), (21, -0.035), (22, 0.036), (23, 0.022), (24, -0.042), (25, -0.01), (26, -0.026), (27, 0.025), (28, -0.044), (29, 0.083), (30, 0.066), (31, -0.019), (32, 0.007), (33, 0.035), (34, -0.078), (35, -0.046), (36, 0.002), (37, -0.02), (38, -0.034), (39, 0.093), (40, 0.087), (41, -0.085), (42, -0.207), (43, 0.013), (44, -0.159), (45, 0.207), (46, -0.086), (47, -0.038), (48, 0.271), (49, -0.206)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.90034294 118 emnlp-2012-Source Language Adaptation for Resource-Poor Machine Translation

Author: Pidong Wang ; Preslav Nakov ; Hwee Tou Ng

2 0.3950637 128 emnlp-2012-Translation Model Based Cross-Lingual Language Model Adaptation: from Word Models to Phrase Models

Author: Shixiang Lu ; Wei Wei ; Xiaoyin Fu ; Bo Xu

3 0.39098665 22 emnlp-2012-Automatically Constructing a Normalisation Dictionary for Microblogs

Author: Bo Han ; Paul Cook ; Timothy Baldwin

Abstract: Microblog normalisation methods often utilise complex models and struggle to differentiate between correctly-spelled unknown words and lexical variants of known words. In this paper, we propose a method for constructing a dictionary of lexical variants of known words that facilitates lexical normalisation via simple string substitution (e.g. tomorrow for tmrw). We use context information to generate possible variant and normalisation pairs and then rank these by string similarity. Highlyranked pairs are selected to populate the dictionary. We show that a dictionary-based approach achieves state-of-the-art performance for both F-score and word error rate on a standard dataset. Compared with other methods, this approach offers a fast, lightweight and easy-to-use solution, and is thus suitable for high-volume microblog pre-processing. 1 Lexical Normalisation A staggering number of short text “microblog” messages are produced every day through social media such as Twitter (Twitter, 2011). The immense volume of real-time, user-generated microblogs that flows through sites has been shown to have utility in applications such as disaster detection (Sakaki et al., 2010), sentiment analysis (Jiang et al., 2011; Gonz a´lez-Ib ´a n˜ez et al., 2011), and event discovery (Weng and Lee, 2011; Benson et al., 2011). However, due to the spontaneous nature of the posts, microblogs are notoriously noisy, containing many non-standard forms e.g., tmrw “tomorrow” and 2day “today” which degrade the performance of — — 421 natural language processing (NLP) tools (Ritter et al., 2010; Han and Baldwin, 2011). To reduce this effect, attempts have been made to adapt NLP tools to microblog data (Gimpel et al., 2011; Foster et al., 2011; Liu et al., 2011b; Ritter et al., 2011). An alternative approach is to pre-normalise non-standard lexical variants to their standard orthography (Liu et al., 2011a; Han and Baldwin, 2011; Xue et al., 2011; Gouws et al., 2011). For example, se u 2morw!!! would be normalised to see you tomorrow! The normalisation approach is especially attractive as a preprocessing step for applications which rely on keyword match or word frequency statistics. For example, earthqu, eathquake, and earthquakeee all attested in a Twitter corpus have the standard form earthquake; by normalising these types to their standard form, better coverage can be achieved for keyword-based methods, and better word frequency estimates can be obtained. In this paper, we focus on the task of lexical normalisation of English Twitter messages, in which out-of-vocabulary (OOV) tokens are normalised to their in-vocabulary (IV) standard form, i.e., a standard form that is in a dictionary. Following other recent work on lexical normalisation (Liu et al., 2011a; Han and Baldwin, 2011; Gouws et al., 2011; Liu et al., 2012), we specifically focus on one-to-one normalisation in which one OOV token is normalised to one IV word. Naturally, not all OOV words in microblogs are lexical variants of IV words: named entities, e.g., — — are prevalent in microblogs, but not all named entities are included in our dictionary. One challenge for lexical normalisation is therefore to disPLraoncge uadgineg Lse oafr tnhineg 2,0 p1a2g Jeosin 42t C1–o4n3f2e,re Jnecjue Iosnla Enmd,p Kiroicraela, M 1e2t–h1o4ds Ju ilny N 20a1tu2r.a ?lc L2a0n1g2ua Agseso Pcrioactieosnsi fnogr a Cnodm Cpoumtaptiuotna tilo Lnianlg Nuaist uircasl tinguish those OOV tokens that require normalisation from those that are well-formed. Recent unsupervised approaches have not attempted to distinguish such tokens from other types of OOV tokens (Cook and Stevenson, 2009; Liu et al., 2011a), limiting their applicability to real-world normalisation tasks. Other approaches (Han and Baldwin, 2011; Gouws et al., 2011) have followed a cascaded approach in which lexical variants are first identified, and then normalised. However, such two-step approaches suffer from poor lexical variant identification performance, which is propagated to the normalisation step. Motivated by the observation that most lexical variants have an unambiguous standard form (especially for longer tokens), and that a lexical variant and its standard form typically occur in similar contexts, in this paper we propose methods for automatically constructing a lexical normalisation dictionary a dictionary whose entries consist — of (lexical variant, standard form) pairs that enables type-based normalisation. Despite the simplicity of this dictionary-based normalisation method, we show it to outperform previously-proposed approaches. This very fast, lightweight solution is suitable for real-time processing of the large volume of streaming microblog data available from Twitter, and offers a simple solution to the lexical variant detection problem that hinders other normalisation methods. Furthermore, this dictionary-based method can be easily integrated with other more-complex normalisation approaches (Liu et al., 2011a; Han and Baldwin, 2011; Gouws et al., 2011) to produce hybrid systems. After discussing related work in Section 2, we present an overview of our dictionary-based approach to normalisation in Section 3. In Sections 4 and 5 we experimentally select the optimised context similarity parameters and string similarity reranking method. We present experimental results on the unseen test data in Section 6, and offer some concluding remarks in Section 7. — 2 Related Work Given a token t, lexical normalisation is the task of finding arg max P(s|t) ∝ arg max P(t| s)P(s), wofh efinred s igs tahreg smtaanxdaPrd(s form, i.e., an aIVx Pw(otr|sd). PSt(asn)-, dardly in lexical normalisation, t is assumed to be an 422 OOV token, relative to a fixed dictionary. In practice, not all OOV tokens should be normalised; i.e., only lexical variants (e.g., tmrw “tomorrow”) should be normalised and tokens that are OOV but otherwise not lexical variants (e.g., iPad “iPad”) should be unchanged. Most work in this area focuses only on the normalisation task itself, oftentimes assuming that the task of lexical variant detection has already been completed. Various approaches have been proposed to estimate the error model, P(t|s). For example, in work on spell-checking, eBl,ril Pl (atn|ds) M. Fooorre e (2000) improve on a standard edit-distance approach by considering multi-character edit operations; Toutanova and Moore (2002) build on this by incorporating phonological information. Li et al. (2006) utilise distributional similarity (Lin, 1998) to correct misspelled search queries. In text message normalisation, Choudhury et al. (2007) model the letter transformations and emissions using a hidden Markov model (Rabiner, 1989). Cook and Stevenson (2009) and Xue et al. (201 1) propose multiple simple error models, each of which captures a particular way in which lexical variants are formed, such as phonetic spelling (e.g., epik “epic”) or clipping (e.g., walkin “walking”). Nevertheless, optimally weighting the various error models in these approaches is challenging. Without pre-categorising lexical variants into different types, Liu et al. (201 1a) collect Google search snippets from carefully-designed queries from which they then extract noisy lexical variant– standard form pairs. These pairs are used to train a conditional random field (Lafferty et al., 2001) to estimate P(t|s) at the character level. One shortcoming eo fP querying a ese cahracrha engine teol. .o Obtanein strhaoirnt-ing pairs is it tends to be costly in terms of time and bandwidth. Here we exploit microblog data directly to derive (lexical variant, standard form) pairs, instead of relying on external resources. In morerecent work, Liu et al. (2012) endeavour to improve the accuracy of top-n normalisation candidates by integrating human cognitive inference, characterlevel transformations and spell checking in their normalisation model. The encouraging results shift the focus to reranking and promoting the correct normalisation to the top-1 position. However, like much previous work on lexical normalisation, this work assumes perfect lexical variant detection. Aw et al. (2006) and Kaufmann and Kalita (2010) consider normalisation as a machine translation task from lexical variants to standard forms using off-theshelf tools. These methods do not assume that lexical variants have been pre-identified; however, these methods do rely on large quantities of labelled training data, which is not available for microblogs. Recently, Han and Baldwin (201 1) and Gouws et al. (201 1) propose two-step unsupervised approaches to normalisation, in which lexical variants are first identified, and then normalised. They approach lexical variant detection by using a context fitness classifier (Han and Baldwin, 2011) or through dictionary lookup (Gouws et al., 2011). However, the lexical variant detection of both meth- ods is rather unreliable, indicating the challenge of this aspect of normalisation. Both of these approaches incorporate a relatively small normalisation dictionary to capture frequent lexical variants with high precision. In particular, Gouws et al. (201 1) produce a small normalisation lexicon based on distributional similarity and string similarity (Lodhi et al., 2002). Our method adopts a similar strategy using distributional/string similarity, but instead of constructing a small lexicon for preprocessing, we build a much wider-coverage normalisation dictionary and opt for a fully lexiconbased end-to-end normalisation approach. In contrast to the normalisation dictionaries of Han and Baldwin (201 1) and Gouws et al. (201 1) which focus on very frequent lexical variants, we focus on moderate frequency lexical variants of a minimum character length, which tend to have unambiguous standard forms; our intention is to produce normalisation lexicons that are complementary to those currently available. Furthermore, we investigate the impact of a variety of contextual and string similarity measures on the quality of the resulting lexicons. In summary, our dictionary-based normalisation ap- proach is a lightweight end-to-end method which performs both lexical variant detection and normalisation, and thus is suitable for practical online preprocessing, despite its simplicity. 423 3 A Lexical Normalisation Dictionary Before discussing our method for creating a normalisation dictionary, we first discuss the feasibility of such an approach. 3.1 Feasibility Dictionary lookup approaches to normalisation have been shown to have high precision but low recall (Han and Baldwin, 2011; Gouws et al., 2011). Frequent (lexical variant, standard form) pairs such as (u, you) are typically included in the dictionaries used by such methods, while less-frequent items such as (g0tta, gotta) are generally omitted. Because of the degree of lexical creativity and large number of non-standard forms observed on Twitter, a wide-coverage normalisation dictionary would be expensive to construct manually. Based on the assumption that lexical variants occur in similar con- texts to their standard forms, however, it should be possible to automatically construct a normalisation dictionary with wider coverage than is currently available. Dictionary lookup is a type-based approach to normalisation, i.e., every token instance of a given type will always be normalised in the same way. However, lexical variants can be ambiguous, e.g., y corresponds to “you” in yeah, y r right! LOL but “why” in AM CONFUSED!!! y you did that? Nevertheless, the relative occurrence of ambiguous lexical variants is small (Liu et al., 2011a), and it has been observed that while shorter variants such as y are often ambiguous, longer variants tend to be unambiguous. For example bthday and 4eva are unlikely to have standard forms other than “birthday” and “forever”, respectively. Therefore, the normalisation lexicons we produce will only contain entries for OOVs with character length greater than a specified threshold, which are likely to have an unambiguous standard form. 3.2 Overview of approach Our method for constructing a normalisation dictio- nary is as follows: Input: Tokenised English tweets 1. Extract (OOV, IV) pairs based on distributional similarity. 2. Re-rank the extracted pairs by string similarity. Output: A list of (OOV, IV) pairs ordered by string similarity; select the top-n pairs for inclusion in the normalisation lexicon. In Step 1, we leverage large volumes of Twitter data to identify the most distributionally-similar IV type for each OOV type. The result of this process is a set of (OOV, IV) pairs, ranked by distributional similarity. The extracted pairs will include (lexical variant, standard form) pairs, such as (tmrw, tomorrow), but will also contain false positives such as (Tusday, Sunday) Tusday is a lexical variant, but its standard form is not “Sunday” and (Youtube, web) Youtube is an OOV named entity, not a lexical variant. Nevertheless, lexical variants are typically formed from their standard forms through regular processes (Thurlow, 2003) e.g., the omission of characters and from this perspective Sunday and web are not plausible standard — — — — — forms for Tusday and Youtube, respectively. In Step 2, we therefore capture this intuition to re-rank the extracted pairs by string similarity. The top-n items in this re-ranked list then form the normalisation lexicon, which is based only on development data. Although computationally-expensive to build, this dictionary can be created offline. Once built, it then offers a very fast approach to normalisation. We can only reliably compute distributional similarity for types that are moderately frequent in a corpus. Nevertheless, many lexical variants are sufficiently frequent to be able to compute distributional similarity, and can potentially make their way into our normalisation lexicon. This approach is not suitable for normalising low-frequency lexical variants, nor is it suitable for shorter lexical variant types which as discussed in Section 3.1 are more likely to have an ambiguous standard form. Nevertheless, previously-proposed normalisation methods that can handle such phenomena also rely in part on a normalisation lexicon. The normalisation lexicons we create can therefore be easily integrated with previous approaches to form hybrid normalisation systems. — — 4 Contextually-similar Pair Generation Our objective is to extract contextually-similar (OOV, IV) pairs from a large-scale collection of mi424 croblog data. Fundamentally, the surrounding words define the primary context, but there are different ways of representing context and different similarity measures we can use, which may influence the quality of generated normalisation pairs. In representing the context, we experimentally explore the following factors: (1) context window size (from 1 to 3 tokens on both sides); (2) n-gram order ofthe context tokens (unigram, bigram, trigram); (3) whether context words are indexed for relative position or not; and (4) whether we use all context tokens, or only IV words. Because high-accuracy linguistic processing tools for Twitter are still under exploration (Liu et al., 2011b; Gimpel et al., 2011; Ritter et al., 2011; Foster et al., 2011), we do not consider richer representations of context, for example, incorporating information about part-of-speech tags or syntax. We also experiment with a number of simple but widely-used geometric and information theoretic distance/similarity measures. In particular, we use Kullback–Leibler (KL) divergence (Kullback and Leibler, 195 1), Jensen–Shannon (JS) divergence (Lin, 1991), Euclidean distance and Cosine distance. We use a corpus of 10 million English tweets to do parameter tuning over, and a larger corpus of tweets in the final candidate ranking. All tweets were collected from September 2010 to January 2011 via the Twitter API.1 From the raw data we extract English tweets using a language identification tool (Lui and Baldwin, 2011), and then apply a simplified Twitter tokeniser (adapted from O’Connor et al. (2010)). We use the Aspell dictionary (v6.06)2 to determine whether a word is IV, and only include in our normalisation dictionary OOV tokens with at least 64 occurrences in the corpus and character length ≥ 4, both of which were determined through empirical 4o,b bsoetrhva otifo wnh. Fcohr w weearceh d OetOeVrm winoedrd t type ginh the corpus, we select the most similar IV type to form (OOV, IV) pairs. To further narrow the search space, we only consider IV words which are morphophonemically similar to the OOV type, follow- ing settings in Han and Baldwin (201 1).3 1http s : / / dev .twitter . com/ docs / st reaming-api /methods 2http : / / aspe l .net / l 3We only consider IV words within an edit distance of 2 or a phonemic edit distance of 1from the OOV type, and we further In order to evaluate the generated pairs, we randomly selected 1000 OOV words from the 10 million tweet corpus. We set up an annotation task on Amazon Mechanical Turk,4 presenting five independent annotators with each word type (with no context) and asking for corrections where appropriate. For instance, given tmrw, the annotators would likely identify it as a non-standard variant of “tomorrow”. For correct OOV words like iPad, on the other hand, we would expect them to leave the word unchanged. If 3 or more of the 5 annotators make the same suggestion (in the form of either a canonical spelling or leaving the word unchanged), we include this in our gold standard for evaluation. In total, this resulted in 351 lexical variants and 282 correct OOV words, accounting for 63.3% of the 1000 OOV words. These 633 OOV words were used as (OOV, IV) pairs for parameter tuning. The remainder of the 1000 OOV words were ignored on the grounds that there was not sufficient consensus amongst the annotators.5 Contextually-similar pair generation aims to include as many correct normalisation pairs as possible. We evaluate the quality of the normalisation pairs using “Cumulative Gain” (CG): XN0 CG = Xreli0 Xi=1 Suppose there are N0 correct generated pairs (oovi, ivi), each of which is weighted by reli0, the frequency of oovi to indicate its relative importance; for example, (thinkin, thinking) has a higher weight than (g0tta, gotta) because thinkin is more frequent than g0tta in our corpus. In this evaluation we don’t consider the position of normalisation pairs, and nor do we penalise incorrect pairs. Instead, we push distinguishing between correct and incorrect pairs into the downstream re-ranking step in which we incorporate string similarity information. Given the development data and CG, we run an exhaustive search of parameter combinations over only consider the top 30% most-frequent of these IV words. 4https : / /www .mturk .com/mturk/welcome 5Note that the objective of this annotation task is to identify lexical variants that have agreed-upon standard forms irrespective of context, as a special case of the more general task of lexical normalisation (where context may or may not play a significant role in the determination of the normalisation). 425 our development corpus. The five best parameter combinations are shown in Table 1. We notice the CG is almost identical for the top combinations. As a context window size of 3 incurs a heavy processing and memory overhead over a size of 2, we use the 3rd-best parameter combination for subsequent experiments, namely: context window of ±2 tokens, teoxkpeenr bigrams, positional index, nadnodw wK oLf divergence as our distance measure. To better understand the sensitivity of the method to each parameter, we perform a post-hoc parameter analysis relative to a default setting (as underlined in Table 2), altering one parameter at a time. The results in Table 2 show that bigrams outperform other n-gram orders by a large margin (note that the evaluation is based on a log scale), and information-theoretic measures are superior to the geometric measures. Furthermore, it also indicates using the positional indexing better captures context. However, there is little to distinguish context modelling with just IV words or all tokens. Similarly, the context window size has relatively little impact on the overall performance, supporting our earlier observation from Table 1. 5 Pair Re-ranking by String Similarity Once the contextually-similar (OOV, IV) pairs are generated using the selected parameters in Section 4, we further re-rank this set of pairs in an attempt to boost morphophonemically-similar pairs like (bananaz, bananas), and penalise noisy pairs like (paninis, beans). Instead of using the small 10 million tweet corpus, from this step onwards, we use a larger corpus of 80 million English tweets (collected over the same period as the development corpus) to develop a larger-scale normalisation dictionary. This is because once pairs are generated, re-ranking based on string comparison is much faster. We only include in the dictionary OOV words with a token frequency > 15 to include more OOV types than in Section 4, and again apply a minimum length cutoff of 4 char- acters. To measure how well our re-ranking method promotes correct pairs and demotes incorrect pairs (including both OOV words that should not be normalised, e.g. (Youtube, web), and incorrect normalRankWindow sizen-gramPositional index?Lex. choiceSim/distance measurelog(CG) 1±32YesAllKL divergence19.571 2 ±±33 2 No All KL divergence 19.562 3 ±±23 2 Yes All KL divergence 19.562 4 ±±32 2 Yes IVs KL divergence 19.561 5 ±±23 2 Yes IVs JS divergence 19.554 ±2 Table 1: The five best parameter combinations in the exhaustive search of parameter combinations Window sizen-gramPositional index?Lexical choiceSimilarity/distance measure ±1 19.3251 19.328Yes 19.328IVs 19.335KL divergence 19.328 ±±21 1199..332275 2 19.571 No 19.263 All 19.328 Euclidean 19.227 ±±32 1199..332287 3 19.324 JS divergence 19.31 1 Cosine 19.170 Table 2: Parameter sensitivity analysis measured as log(CG) for correctly-generated pairs. We tune one parameter at a time, using the default (underlined) setting for other parameters; the non-exhaustive best-performing setting in each case is indicated in bold. isations for lexical variants, e.g. (bcuz, cause)), we modify our evaluation metric from Section 4 to evaluate the ranking at different points, using Discounted Cumulative Gain (DCG@N: Jarvelin and Kekalainen (2002)): DCG@N = rel1+XiN=2logr2el(i ) where reli again represents the frequency of the OOV, but it can be gain (a positive number) or loss (a negative number), depending on whether the ith pair is correct or incorrect. Because we also expect correct pairs to be ranked higher than incorrect pairs, DCG@N takes both factors into account. Given the generated pairs and the evaluation metric, we first consider three baselines: no re-ranking (i.e., the final ranking is that of the contextual similarity scores), and re-rankings of the pairs based on the frequencies of the OOVs in the Twitter corpus, and the IV unigram frequencies in the Google Web 1T corpus (Brants and Franz, 2006) to get less-noisy frequency estimates. We also compared a variety of re-rankings based on a number of string similarity measures that have been previously considered in normalisation work (reviewed in Section 2). We experiment with standard edit distance (Levenshtein, 1966), edit distance over double metaphone codes (phonetic edit distance: (Philips, 2000)), longest common subsequence ratio over the consonant edit distance of the paired words (hereafter, denoted as 426 consonant edit distance: (Contractor et al., 2010)), and a string subsequence kernel (Lodhi et al., 2002). In Figure 1, we present the DCG@N results for each of our ranking methods at different rank cutoffs. Ranking by OOV frequency is motivated by the assumption that lexical variants are frequently used by social media users. This is confirmed by our findings that lexical pairs like (goin, going) and (nite, night) are at the top of the ranking. However, many proper nouns and named entities are also used frequently and ranked at the top, mixed with lexical variants like (Facebook, speech) and (Youtube, web). In ranking by IV word frequency, we assume the lexical variants are usually derived from frequently-used IV equivalents, e.g. (abou, about). However, many less-frequent lexical variant types have high-frequency (IV) normalisations. For instance, the highest-frequency IV word the has more than 40 OOV lexical variants, such as tthe and thhe. These less-frequent types occupy the top positions, reducing the cumulative gain. Compared with these two baselines, ranking by default contextual similarity scores delivers promising results. It successfully ranks many more intuitive normalisation pairs at the top, such as (2day, today) and (wknd, weekend), but also ranks some incorrect pairs highly, such as (needa, gotta). The string similarity-based methods perform better than our baselines in general. Through manual analysis, we found that standard edit distance ranking is fairly accurate for lexical variants with low edit distance to their standard forms, but fails to identify heavily-altered variants like (tmrw, tomorrow). Consonant edit distance is similar to standard edit distance, but places many longer words at the top of the ranking. Edit distance over double metaphone codes (phonetic edit distance) performs particularly well for lexical variants that include character repetitions commonly used for emphasis on Twitter because such repetitions do not typically alter the phonetic codes. Compared with the other methods, the string subsequence kernel delivers encouraging results. It measures common character subsequences of length n between (OOV, IV) pairs. Because it is computationally expensive to calculate similarity for larger n, we choose n=2, following Gouws et al. (201 1). As N (the lexicon size cut-off) increases, the performance drops more slowly than the other meth— — ods. Although this method fails to rank heavilyaltered variants such as (4get,forget) highly, it typically works well for longer words. Given that we focus on longer OOVs (specifically those longer than 4 characters), this ultimately isn’t a great handicap. 6 Evaluation Given the re-ranked pairs from Section 5, here we apply them to a token-level normalisation task using the normalisation dataset of Han and Baldwin (201 1). 6.1 Metrics We evaluate using the standard evaluation metrics of precision (P), recall (R) and F-score (F) as detailed below. We also consider the false alarm rate (FA) and word error rate (WER), also as shown below. FA measures the negative effects of applying normalisation; a good approach to normalisation should not (incorrectly) normalise tokens that are already in their standard form and do not require normalisation.6 WER, like F-score, shows the overall benefits of normalisation, but unlike F-score, measures how many token-level edits are required for the output to be the same as the ground truth data. In general, dictionaries with a high F-score/low WER and low FA 6FA + P ≤ 1because some lexical variants might be incorrectly Ano +rm Pa ≤lise 1d b. 427 are preferable. P = R= F = FA = WER = # cor#re nctolrym naolrismedal tioskeden toskens # to ckoernresc rtelyqu niori nmga nloisremda tloiskaetniosn P2P +R R # inco#rr encotrlmya nliosremda tloikseedns tokens # token edits n#ee adlletd o akfetnesr normalisation 6.2 Results We select the three best re-ranking methods, and best cut-off N for each method, based on the highest DCG@N value for a given method over the development data, as presented in Figure 1. Namely, they are string subsequence kernel (S-dict, N=40,000), double metaphone edit distance (DMdict, N=10,000) and default contextual similarity without re-ranking (C-dict, N=10,000).7 We evaluate each of the learned dictionaries in Table 3. We also compare each dictionary with the performance of the manually-constructed Internet slang dictionary (HB-dict) used by Han and Baldwin (201 1), the small automatically-derived dictionary of Gouws et al. (201 1) (GHM-dict), and combinations of the different dictionaries. In addition, the contribution of these dictionaries in hybrid normalisation approaches is also presented, in which we first normalise OOVs using a given dictionary (combined or otherwise), and then apply the normalisation method of Gouws et al. (201 1) based on consonant edit distance (GHM-norm), or the approach of Han and Baldwin (201 1) based on the summation of many unsupervised approaches (HB-norm), to the remaining OOVs. Results are shown in Table 3, and discussed below. 6.2.1 Individual Dictionaries Overall, the individual dictionaries derived by the re-ranking methods (DM-dict, S-dict) perform bet- 7We also experimented with combining ranks using Mean Reciprocal Rank. However, the combined rank didn’t improve performance on the development data. We plan to explore other ranking aggregation methods in future work. 1 3 5 7 9 11 31 51 71 91 N cut−offs Figure 1: Re-ranking based on different string similarity methods. ter than that based on contextual similarity (C-dict) in terms of precision and false alarm rate, indicating the importance of re-ranking. Even though C-dict delivers higher recall indicating that many lexical variants are correctly normalised this is offset by its high false alarm rate, which is particularly undesirable in normalisation. Because S-dict has better performance than DM-dict in terms of both F-score and WER, and a much lower false alarm rate than C-dict, subsequent results are presented using S-dict only. — — Both HB-dict and GHM-dict achieve better than 90% precision with moderate recall. Compared to these methods, S-dict is not competitive in terms of either precision or recall. This result seems rather discouraging. However, considering that S-dict is an automatically-constructed dictionary targeting lexical variants of varying frequency, it is not surprising that the precision is worse than that of HB-dict which is manually-constructed and GHM-dict which includes entries only for more-frequent OOVs for which distributional similarity is more accurate. Additionally, the recall of S-dict is hampered by the — — — 428 restriction on lexical variant token length of 4 characters. 6.2.2 Combined Dictionaries Next we look to combining HB-dict, GHM-dict and S-dict. In combining the dictionaries, a given OOV word can be listed with different standard forms in different dictionaries. In such cases we use the following preferences for dictionaries motivated by our confidence in the normalisation pairs — of the dictionaries to resolve conflicts: HB-dict > GHM-dict > S-dict. When we combine dictionaries in the second section of Table 3, we find that they contain complementary information: in each case the recall and F-score are higher for the combined dictionary than any of the individual dictionaries. The combination of HB-dict+GHM-dict produces only a small improvement in terms of F-score over HBdict (the better-performing dictionary) suggesting that, as claimed, HB-dict and GHM-dict share many frequent normalisation pairs. HB-dict+S-dict and GHM-dict+S-dict, on the other hand, improve sub— MethodPrecisionRecallF-ScoreFalse AlarmWord Error Rate C-dict0.4740.2180.2990.2980.103 DM-dict S-dict HB-dict GHM-dict 0.727 0.700 0.915 0.982 0.106 0.179 0.435 0.319 0.185 0.285 0.590 0.482 0.145 0.162 0.048 0.000 0.102 0.097 0.066 0.076 HB-dict+S-dict0.8400.6010.7010.0900.052 GHM-dict+S-dict HB-dict+GHM-dict HB-dict+GHM-dict+S-dict 0.863 0.920 0.847 0.498 0.465 0.630 0.632 0.618 0.723 0.072 0.045 0.086 0.061 0.063 0.049 GHM-dict+GHM-norm0.3380.5780.4270.4580.135 HB-dict+GHM-dict+S-dict+GHM-norm HB-dict+HB-norm HB-dict+GHM-dict+S-dict+HB-norm 0.406 0.515 0.527 0.715 0.771 0.789 0.518 0.618 0.632 0.468 0.332 0.332 0.124 0.081 0.079 Table 3: Normalisation results using our derived dictionaries (contextual similarity (C-dict); double metaphone rendering (DM-dict); string subsequence kernel scores (S-dict)), the dictionary of Gouws et al. (201 1) (GHM-dict), the Internet slang dictionary (HB-dict) from Han and Baldwin (201 1), and combinations of these dictionaries. In addition, we combine the dictionaries with the normalisation method of Gouws et al. (201 1) (GHM-norm) and the combined unsupervised approach of Han and Baldwin (201 1) (HB-norm). stantially over HB-dict and GHM-dict, respectively, indicating that S-dict contains markedly different entries to both HB-dict and GHM-dict. The best Fscore and WER are obtained using the combination of all three dictionaries, HB-dict+GHM-dict+S-dict. Furthermore, the difference between the results using HB-dict+GHM-dict+S-dict and HB-dict+GHMdict is statistically significant (p < 0.01), based on the computationally-intensive Monte Carlo method of Yeh (2000), demonstrating the contribution of Sdict. 6.2.3 Hybrid Approaches The methods of Gouws et al. (201 1) (i.e. GHM-dict+GHM-norm) and Han and Baldwin (201 1) (i.e. HB-dict+HB-norm) have lower precision and higher false alarm rates than the dictionarybased approaches; this is largely caused by lexical variant detection errors.8 Using all dictionaries in combination with these methods HB-dict+GHM-dict+S-dict+GHM-norm and HBdict+GHM-dict+S-dict+HB-norm gives some improvements, but the false alarm rates remain high. Despite the limitations of a pure dictionary-based approach to normalisation discussed in Section 3.1 the current best practical approach to normal— — — — 8Here we report results that do not assume perfect detection of lexical variants, unlike the original published results in each case. 429 Error typeOOVDSitcat.ndard fGoromld (a) pluralsplayeplayersplayer (b) negation unlike like dislike (c) possessives anyones anyone anyone ’s (d) correct OOVs iphone phone iphone (e) test data errors durin during durin (f) ambiguity siging signing singing Table 4: Error types in the combined dictionary (HBdict+GHM-dict+S-dict) isation is to use a lexicon, combining hand-built and automatically-learned normalisation dictionaries. 6.3 Discussion and Error Analysis We first manually analyse the errors in the combined dictionary (HB-dict+GHM-dict+S-dict) and give examples of each error type in Table 4. The most frequent word errors are caused by slight morphologi- cal variations, including plural forms (a), negations (b), possessive cases (c), and OOVs that are correct and do not require normalisation (d). In addition, we also notice some missing annotations where lexical variants are skipped by human annotations but captured by our method (e). Ambiguity (f) definitely exists in longer OOVs, however, these cases do not appear to have a strong negative impact on the normalisation performance. An example of a remainLength cut-off (N)#VariantsPrecisionRecall (≥ N)Recall (all)False Alarm ≥45560.700Rec0al.l3 8(≥1 N)0.1790.162 ≥≥54 382 0.814 0.471 0.152 0.122 ≥≥65 254 0.804 0.484 0.104 0.131 ≥≥76 138 0.793 0.471 0.055 0.122 ≥71380.7930.4710.0550.122 Table 5: S-dict normalisation results broken down according to OOV token length. Recall is presented both over the subset of instances of length ≥ N in the data (“Recall (≥ N)”), and over the entirety of the dataset (“Recall (all)”); “su#bVsaertia onftis n” sitsa tnhcee snu omfb leenrg othf t≥ok Nen iinns tthaenc deast ao f( “tRhee cinadllic (a≥ted N length idn o othveer rt tehset d eanttaisreetty. ing miscellaneous error is bday “birthday”, which is mis-normalised as day. To further study the influence of OOV word length relative to the normalisation performance, we conduct a fine-grained analysis of the performance of the derived dictionary (S-dict) in Table 5, broken down across different OOV word lengths. The results generally support our hypothesis that our method works better for longer OOV words. The derived dictionary is much more reliable for longer tokens (length 5, 6, and 7 characters) in terms of precision and false alarm. Although the recall is relatively modest, in the future we intend to improve recall by mining more normalisation pairs from larger collections of microblog data. 7 Conclusions and Future Work In this paper, we describe a method for automatically constructing a normalisation dictionary that supports normalisation of microblog text through direct substitution of lexical variants with their standard forms. After investigating the impact of different distributional and string similarity methods on the quality of the dictionary, we present experimental results on a standard dataset showing that our proposed methods acquire high quality (lexical variant, standard form) pairs, with reasonable coverage, and achieve state-of-the-art end-toend lexical normalisation performance on a realworld token-level task. Furthermore, this dictionarylookup method combines the detection and normalisation of lexical variants into a simple, lightweight solution which is suitable for processing of highvolume microblog feeds. In the future, we intend to improve our dictionary by leveraging the constantly-growing volume of microblog data, and considering alternative ways to combine distributional and string similarity. In addi430 tion to direct evaluation, we also want to explore the benefits of applying normalisation for downstream social media text processing applications, e.g. event detection. Acknowledgements We would like to thank the three anonymous reviewers for their insightful comments, and Stephan Gouws for kindly sharing his data and discussing his work. NICTA is funded by the Australian government as represented by Department of Broadband, Communication and Digital Economy, and the Australian Research Council through the ICT centre of Excellence programme. References AiTi Aw, Min Zhang, Juan Xiao, and Jian Su. 2006. A phrase-based statistical model for SMS text normalization. In Proceedings of COLING/ACL 2006, pages 33–40, Sydney, Australia. Edward Benson, Aria Haghighi, and Regina Barzilay. 2011. Event discovery in social media feeds. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 389–398, Portland, Oregon, USA. Thorsten Brants and Alex Franz. 2006. Web 1T 5-gram Version 1. Eric Brill and Robert C. Moore. 2000. An improved error model for noisy channel spelling correction. In Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, pages 286–293, Hong Kong. Monojit Choudhury, Rahul Saraf, Vijit Jain, Animesh Mukherjee, Sudeshna Sarkar, and Anupam Basu. 2007. Investigation and modeling of the structure of texting language. International Journal on Document Analysis and Recognition, 10: 157–174. Danish Contractor, Tanveer A. Faruquie, and L. Venkata Subramaniam. 2010. Unsupervised cleansing of noisy text. In Proceedings of the 23rd International Conference on Computational Linguistics (COLING 2010), pages 189–196, Beijing, China. Paul Cook and Suzanne Stevenson. 2009. An unsu- pervised model for text message normalization. In CALC ’09: Proceedings of the Workshop on Computational Approaches to Linguistic Creativity, pages 71– 78, Boulder, USA. Jennifer Foster, O¨zlem C ¸etinoglu, Joachim Wagner, Joseph L. Roux, Stephen Hogan, Joakim Nivre, Deirdre Hogan, and Josef van Genabith. 2011. #hardtoparse: POS Tagging and Parsing the Twitterverse. In Analyzing Microtext: Papers from the 2011 AAAI Workshop, volume WS-1 1-05 of AAAI Workshops, pages 20–25, San Francisco, CA, USA. Kevin Gimpel, Nathan Schneider, Brendan O’Connor, Dipanjan Das, Daniel Mills, Jacob Eisenstein, Michael Heilman, Dani Yogatama, Jeffrey Flanigan, and Noah A. Smith. 2011. Part-of-speech tagging for Twitter: Annotation, features, and experiments. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 42–47, Portland, Oregon, USA. Roberto Gonz a´lez-Ib ´a n˜ez, Smaranda Muresan, and Nina Wacholder. 2011. Identifying sarcasm in Twitter: a closer look. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 581–586, Portland, Oregon, USA. Stephan Gouws, Dirk Hovy, and Donald Metzler. 2011. Unsupervised mining of lexical variants from noisy text. In Proceedings of the First workshop on Unsupervised Learning in NLP, pages 82–90, Edinburgh, Scotland, UK. Bo Han and Timothy Baldwin. 2011. Lexical normalisation of short text messages: Makn sens a #twitter. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 368–378, Portland, Oregon, USA. K. Jarvelin and J. Kekalainen. 2002. Cumulated gainbased evaluation of IR techniques. ACM Transactions on Information Systems, 20(4). Long Jiang, Mo Yu, Ming Zhou, Xiaohua Liu, and Tiejun Zhao. 2011. Target-dependent Twitter sentiment classification. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 151–160, Portland, Oregon, USA. Joseph Kaufmann and Jugal Kalita. 2010. Syntactic normalization of Twitter messages. In International Con431 ference on Natural Language Processing, Kharagpur, India. S. Kullback and R. A. Leibler. 1951. On information and sufficiency. Annals of Mathematical Statistics, 22:49– 86. John D. Lafferty, Andrew McCallum, and Fernando C. N. Pereira. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings ofthe Eighteenth International Conference on Machine Learning, pages 282–289, San Francisco, CA, USA. Vladimir I. Levenshtein. 1966. Binary codes capable of correcting deletions, insertions, and reversals. Soviet Physics Doklady, 10:707–710. Mu Li, Yang Zhang, Muhua Zhu, and Ming Zhou. 2006. Exploring distributional similarity based models for query spelling correction. In Proceedings of COLING/ACL 2006, pages 1025–1032, Sydney, Australia. Jianhua Lin. 1991. Divergence measures based on the shannon entropy. IEEE Transactions on Information Theory, 37(1): 145–151. Dekang Lin. 1998. Automatic retrieval and clustering of similar words. In Proceedings of the 36th Annual Meeting of the ACL and 1 International Con7th ference on Computational Linguistics (COLING/ACL98), pages 768–774, Montreal, Quebec, Canada. Fei Liu, Fuliang Weng, Bingqing Wang, and Yang Liu. 2011a. Insertion, deletion, or substitution? normalizing text messages without pre-categorization nor supervision. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 71–76, Portland, Oregon, USA. Xiaohua Liu, Shaodian Zhang, Furu Wei, and Ming Zhou. 2011b. Recognizing named entities in tweets. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies (ACL-HLT 2011), pages 359–367, Portland, Oregon, USA. Fei Liu, Fuliang Weng, and Xiao Jiang. 2012. A broadcoverage normalization system for social media language. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (ACL 2012), Jeju, Republic of Korea. Huma Lodhi, Craig Saunders, John Shawe-Taylor, Nello Cristianini, and Chris Watkins. 2002. Text classification using string kernels. J. Mach. Learn. Res., 2:419– 444. Marco Lui and Timothy Baldwin. 2011. Cross-domain feature selection for language identification. In Proceedings of the 5th International Joint Conference on Natural Language Processing (IJCNLP 2011), pages 553–561, Chiang Mai, Thailand. Brendan O’Connor, Michel Krieger, and David Ahn. 2010. TweetMotif: Exploratory search and topic summarization for Twitter. In Proceedings of the 4th International Conference on Weblogs and Social Media (ICWSM 2010), pages 384–385, Washington, USA. Lawrence Philips. 2000. The double metaphone search algorithm. C/C++ Users Journal, 18:38–43. Lawrence R. Rabiner. 1989. A tutorial on hidden Markov models and selected applications in speech recognition. Proceedings of the IEEE, 77(2):257–286. Alan Ritter, Colin Cherry, and Bill Dolan. 2010. Unsupervised modeling of Twitter conversations. In Proceedings of Human Language Technologies: The 11th Annual Conference of the North American Chap- ter of the Association for Computational Linguistics (NAACL-HLT 2010), pages 172–180, Los Angeles, USA. Alan Ritter, Sam Clark, Mausam, and Oren Etzioni. 2011. Named entity recognition in tweets: An experimental study. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing (EMNLP 2011), pages 1524–1534, Edinburgh, Scotland, UK. Takeshi Sakaki, Makoto Okazaki, and Yutaka Matsuo. 2010. Earthquake shakes Twitter users: real-time event detection by social sensors. In Proceedings of the 19th International Conference on the World Wide Web (WWW 2010), pages 851–860, Raleigh, North Carolina, USA. Crispin Thurlow. 2003. Generation txt? The sociolinguistics of young people’s text-messaging. Discourse Analysis Online, 1(1). Kristina Toutanova and Robert C. Moore. 2002. Pronunciation modeling for improved spelling correction. In Proceedings of the 40th Annual Meeting of the ACL and 3rd Annual Meeting of the NAACL (ACL-02), pages 144–15 1, Philadelphia, USA. Official Blog Twitter. 2011. 200 million tweets per day. Retrived at August 17th, 2011. Jianshu Weng and Bu-Sung Lee. 2011. Event detection in Twitter. In Proceedings of the 5th International Conference on Weblogs and Social Media (ICWSM 2011), Barcelona, Spain. Zhenzhen Xue, Dawei Yin, and Brian D. Davison. 2011. Normalizing microtext. In Proceedings of the AAAI11 Workshop on Analyzing Microtext, pages 74–79, San Francisco, USA. Alexander Yeh. 2000. More accurate tests for the statistical significance of result differences. In Proceedings of the 23rd International Conference on Computational Linguistics (COLING 2010), pages 947–953, Saarbr¨ ucken, Germany. 432

4 0.3478429 9 emnlp-2012-A Sequence Labelling Approach to Quote Attribution

Author: Timothy O'Keefe ; Silvia Pareti ; James R. Curran ; Irena Koprinska ; Matthew Honnibal

Abstract: Quote extraction and attribution is the task of automatically extracting quotes from text and attributing each quote to its correct speaker. The present state-of-the-art system uses gold standard information from previous decisions in its features, which, when removed, results in a large drop in performance. We treat the problem as a sequence labelling task, which allows us to incorporate sequence features without using gold standard information. We present results on two new corpora and an augmented version of a third, achieving a new state-of-the-art for systems using only realistic features.

5 0.31826302 26 emnlp-2012-Building a Lightweight Semantic Model for Unsupervised Information Extraction on Short Listings

Author: Doo Soon Kim ; Kunal Verma ; Peter Yeh

Abstract: Short listings such as classified ads or product listings abound on the web. If a computer can reliably extract information from them, it will greatly benefit a variety of applications. Short listings are, however, challenging to process due to their informal styles. In this paper, we present an unsupervised information extraction system for short listings. Given a corpus of listings, the system builds a semantic model that represents typical objects and their attributes in the domain of the corpus, and then uses the model to extract information. Two key features in the system are a semantic parser that extracts objects and their attributes and a listing-focused clustering module that helps group together extracted tokens of same type. Our evaluation shows that the , semantic model learned by these two modules is effective across multiple domains.

6 0.30156952 75 emnlp-2012-Large Scale Decipherment for Out-of-Domain Machine Translation

7 0.27582094 86 emnlp-2012-Locally Training the Log-Linear Model for SMT

8 0.27187178 99 emnlp-2012-On Amortizing Inference Cost for Structured Prediction

9 0.27152795 122 emnlp-2012-Syntactic Surprisal Affects Spoken Word Duration in Conversational Contexts

10 0.24958539 114 emnlp-2012-Revisiting the Predictability of Language: Response Completion in Social Media

11 0.2124429 54 emnlp-2012-Forced Derivation Tree based Model Training to Statistical Machine Translation

12 0.20968053 74 emnlp-2012-Language Model Rest Costs and Space-Efficient Storage

13 0.19727844 131 emnlp-2012-Unified Dependency Parsing of Chinese Morphological and Syntactic Structures

14 0.19192648 58 emnlp-2012-Generalizing Sub-sentential Paraphrase Acquisition across Original Signal Type of Text Pairs

15 0.18910059 39 emnlp-2012-Enlarging Paraphrase Collections through Generalization and Instantiation

16 0.18675956 108 emnlp-2012-Probabilistic Finite State Machines for Regression-based MT Evaluation

17 0.18368512 31 emnlp-2012-Cross-Lingual Language Modeling with Syntactic Reordering for Low-Resource Speech Recognition

18 0.18105242 35 emnlp-2012-Document-Wide Decoding for Phrase-Based Statistical Machine Translation

19 0.16931911 27 emnlp-2012-Characterizing Stylistic Elements in Syntactic Structure

20 0.16294971 121 emnlp-2012-Supervised Text-based Geolocation Using Language Models on an Adaptive Grid

similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(2, 0.022), (11, 0.014), (16, 0.025), (34, 0.074), (45, 0.02), (50, 0.34), (60, 0.098), (63, 0.045), (64, 0.014), (65, 0.018), (70, 0.022), (74, 0.039), (76, 0.046), (79, 0.017), (80, 0.013), (81, 0.011), (86, 0.03), (95, 0.019)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.697855 118 emnlp-2012-Source Language Adaptation for Resource-Poor Machine Translation

Author: Pidong Wang ; Preslav Nakov ; Hwee Tou Ng

2 0.49147478 136 emnlp-2012-Weakly Supervised Training of Semantic Parsers

Author: Jayant Krishnamurthy ; Tom Mitchell

Abstract: We present a method for training a semantic parser using only a knowledge base and an unlabeled text corpus, without any individually annotated sentences. Our key observation is that multiple forms ofweak supervision can be combined to train an accurate semantic parser: semantic supervision from a knowledge base, and syntactic supervision from dependencyparsed sentences. We apply our approach to train a semantic parser that uses 77 relations from Freebase in its knowledge representation. This semantic parser extracts instances of binary relations with state-of-theart accuracy, while simultaneously recovering much richer semantic structures, such as conjunctions of multiple relations with partially shared arguments. We demonstrate recovery of this richer structure by extracting logical forms from natural language queries against Freebase. On this task, the trained semantic parser achieves 80% precision and 56% recall, despite never having seen an annotated logical form.

3 0.39600417 123 emnlp-2012-Syntactic Transfer Using a Bilingual Lexicon

Author: Greg Durrett ; Adam Pauls ; Dan Klein

Abstract: We consider the problem of using a bilingual dictionary to transfer lexico-syntactic information from a resource-rich source language to a resource-poor target language. In contrast to past work that used bitexts to transfer analyses of specific sentences at the token level, we instead use features to transfer the behavior of words at a type level. In a discriminative dependency parsing framework, our approach produces gains across a range of target languages, using two different lowresource training methodologies (one weakly supervised and one indirectly supervised) and two different dictionary sources (one manually constructed and one automatically constructed).

4 0.39308187 138 emnlp-2012-Wiki-ly Supervised Part-of-Speech Tagging

Author: Shen Li ; Joao Graca ; Ben Taskar

Abstract: Despite significant recent work, purely unsupervised techniques for part-of-speech (POS) tagging have not achieved useful accuracies required by many language processing tasks. Use of parallel text between resource-rich and resource-poor languages is one source ofweak supervision that significantly improves accuracy. However, parallel text is not always available and techniques for using it require multiple complex algorithmic steps. In this paper we show that we can build POS-taggers exceeding state-of-the-art bilingual methods by using simple hidden Markov models and a freely available and naturally growing resource, the Wiktionary. Across eight languages for which we have labeled data to evaluate results, we achieve accuracy that significantly exceeds best unsupervised and parallel text methods. We achieve highest accuracy reported for several languages and show that our . approach yields better out-of-domain taggers than those trained using fully supervised Penn Treebank.

5 0.39297038 70 emnlp-2012-Joint Chinese Word Segmentation, POS Tagging and Parsing

Author: Xian Qian ; Yang Liu

Abstract: In this paper, we propose a novel decoding algorithm for discriminative joint Chinese word segmentation, part-of-speech (POS) tagging, and parsing. Previous work often used a pipeline method Chinese word segmentation followed by POS tagging and parsing, which suffers from error propagation and is unable to leverage information in later modules for earlier components. In our approach, we train the three individual models separately during training, and incorporate them together in a unified framework during decoding. We extend the CYK parsing algorithm so that it can deal with word segmentation and POS tagging features. As far as we know, this is the first work on joint Chinese word segmentation, POS tagging and parsing. Our experimental results on Chinese Tree Bank 5 corpus show that our approach outperforms the state-of-the-art pipeline system. –

6 0.39181757 92 emnlp-2012-Multi-Domain Learning: When Do Domains Matter?

7 0.3912389 18 emnlp-2012-An Empirical Investigation of Statistical Significance in NLP

8 0.3901518 71 emnlp-2012-Joint Entity and Event Coreference Resolution across Documents

9 0.38994518 39 emnlp-2012-Enlarging Paraphrase Collections through Generalization and Instantiation

10 0.389175 42 emnlp-2012-Entropy-based Pruning for Phrase-based Machine Translation

11 0.38905793 109 emnlp-2012-Re-training Monolingual Parser Bilingually for Syntactic SMT

12 0.38877174 24 emnlp-2012-Biased Representation Learning for Domain Adaptation

13 0.38875833 23 emnlp-2012-Besting the Quiz Master: Crowdsourcing Incremental Classification Games

14 0.38796264 107 emnlp-2012-Polarity Inducing Latent Semantic Analysis

15 0.38783926 127 emnlp-2012-Transforming Trees to Improve Syntactic Convergence

16 0.38769564 108 emnlp-2012-Probabilistic Finite State Machines for Regression-based MT Evaluation

17 0.38706186 135 emnlp-2012-Using Discourse Information for Paraphrase Extraction

18 0.38680407 14 emnlp-2012-A Weakly Supervised Model for Sentence-Level Semantic Orientation Analysis with Multiple Experts

19 0.38649216 52 emnlp-2012-Fast Large-Scale Approximate Graph Construction for NLP

20 0.38590118 45 emnlp-2012-Exploiting Chunk-level Features to Improve Phrase Chunking