acl acl2011 acl2011-163 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Thomas Mueller ; Hinrich Schuetze
Abstract: We present a class-based language model that clusters rare words of similar morphology together. The model improves the prediction of words after histories containing outof-vocabulary words. The morphological features used are obtained without the use of labeled data. The perplexity improvement compared to a state of the art Kneser-Ney model is 4% overall and 81% on unknown histories.
Reference: text
sentIndex sentText sentNum sentScore
1 The model improves the prediction of words after histories containing outof-vocabulary words. [sent-2, score-0.163]
2 The morphological features used are obtained without the use of labeled data. [sent-3, score-0.251]
3 The perplexity improvement compared to a state of the art Kneser-Ney model is 4% overall and 81% on unknown histories. [sent-4, score-0.456]
4 1 Introduction One of the challenges in statistical language modeling are words that appear in the recognition task at hand, but not in the training set, so called outof-vocabulary (OOV) words. [sent-5, score-0.085]
5 We present a novel approach based on morphological classes to handling OOV . [sent-7, score-0.346]
6 Previous work on morphological classes in English has not been able to show noticeable improvements in perplexity. [sent-9, score-0.346]
7 Our model improves perplexity of a Kneser-Ney (KN) model for English by 4%, the largest improvement of a state-of-the-art model for English due to morphological modeling that we are aware of. [sent-12, score-0.705]
8 A class-based language model groups words into classes and replaces the word transition probability by a class transition probability and a word emission probability: P(w3 |w1w2) = P(c3 |c1c2) · P(w3 |c3) . [sent-13, score-0.417]
9 (1998) showed that context clustering works better than clusters based on part-of-speech tags. [sent-18, score-0.19]
10 However, since the context of an OOV word is unknown and it therefore cannot be assigned to a cluster, OOV words are as much a problem to a context-based class model as to a word model. [sent-19, score-0.274]
11 That is why we use non-distributional features features like morphological suffixes that only depend on the shape of the word itself to design a new class-based model that can naturally integrate unknown words. [sent-20, score-0.655]
12 In related work, factored language models (Bilmes and Kirchhoff, 2003) were proposed to make use of morphological information in highly inflecting languages such as Finnish (Creutz et al. [sent-21, score-0.398]
13 , 2004) or compounding languages like German (Berton et al. [sent-25, score-0.035]
14 The main idea is to replace words by sequences of factors or features and to apply statistical language modeling to the resulting factor sequences. [sent-27, score-0.085]
15 If, for example, words were segmented into morphemes, an unknown word would be split into an unseen sequence, which could be recognized using discounting techniques. [sent-28, score-0.187]
16 the stem, is unknown to the system, the fundamental problem remains unsolved. [sent-31, score-0.153]
17 – – Our class-based model uses a number of features that have not been used in factored models (e. [sent-32, score-0.124]
18 , shape and length features) and achieves in con– trast to factored models English. [sent-34, score-0.144]
19 – good perplexity gains for Proceedings ofP thoer t4l9atnhd A, Onrnuegaoln M,e Jeuntineg 19 o-f2 t4h,e 2 A0s1s1o. [sent-35, score-0.253]
20 2 Morphological Features The feature vector of a word consists of four parts that represent information about suffixes, capitalization, special characters and word length. [sent-38, score-0.153]
21 For the suffix group, we define a binary feature for each of the 100 most frequent suffixes learned on the training corpus by the Reports algorithm (Keshava, 2006), a general purpose unsupervised morphology learning algorithm. [sent-39, score-0.3]
22 One additional binary feature is used for all other suffixes learned by Reports, including the empty suffix. [sent-40, score-0.174]
23 The feature groups capitalization and special characters are motivated by the analysis shown in Table 2. [sent-41, score-0.288]
24 The table shows that most OOV words (f = 0) are numbers (CD), names (NP), and nouns and adjectives (NN, NNS, JJ). [sent-43, score-0.158]
25 This distribution is similar to hapax legomena (f = 1), but different from the POS distribution of all tokens. [sent-44, score-0.038]
26 Capitalization and special character features are of obvious utility in identifying the POS classes NP and CD since names in English are usually capitalized and numbers are written with digits and special characters such as comma and period. [sent-45, score-0.334]
27 The four parts of the vector (suffixes, capitalization, special characters, length) are weighted equally by normalizing the subvector of each subgroup to unit length. [sent-51, score-0.052]
28 525 We designed the four feature groups to group word types to either resemble POS classes or to in- duce an even finer sub-partitioning. [sent-52, score-0.248]
29 Unsupervised POS clustering is a hard task in English and it is virtually impossible if a word’s context (which is not available for OOV items) is not taken into account. [sent-53, score-0.135]
30 But as our analysis in Table 2 shows, part of the benefit of morphological analysis for OOVs comes from an appropriate treatment of names and numbers. [sent-55, score-0.291]
31 The suffix feature group is useful for categorizing OOV nouns and adjectives because there are very few irregular morphemes like “ren” in children in English and OOV words are likely to be regular words. [sent-56, score-0.303]
32 So even though morphological learning based on the limited information we use is not possible in general, it can be partially solved for the special case of OOV words. [sent-57, score-0.303]
33 That is why we use a frequency-based approach and only include words below a certain token frequency threshold θ in the clustering process. [sent-64, score-0.269]
34 A second motivation is that the contexts of low frequency words are more similar to the expected contexts of OOV words. [sent-65, score-0.069]
35 Given a training corpus, all words with a fre- ing set frequencies f oinfg gaslle tvefrrebq PuOenSc tags. [sent-66, score-0.034]
36 V* consists quency below the threshold θ are partitioned into k clusters using the bisecting k-means algorithm (Steinbach et al. [sent-68, score-0.155]
37 The cluster of an OOV word w can be defined as the cluster whose centroid is closest to the feature vector of w. [sent-70, score-0.227]
38 The formerly removed high-frequency words are added as singleton clusters to produce a complete clustering. [sent-71, score-0.175]
39 However, OOV words can only be assigned to the original k-means clusters. [sent-72, score-0.034]
40 Over this clustering a classbased trigram model can be defined, as introduced by Brown et al. [sent-73, score-0.239]
41 The word transition probability of such a model is given by equation 1, where ci denotes the cluster of the word wi. [sent-75, score-0.201]
42 The class transition probability P(c3 |c1c2) is estimated using the unsmoothed maximum |lcikelihood estimate. [sent-76, score-0.096]
43 The emission probability is defined as follows: wP(her3|c (w)=isthǫ(e1f−reqǫu)ePncwy∈(cwo3fc)w( inthfeθc tr(≥awinc3()inw>=g3)sθ0e>t. [sent-77, score-0.045]
44 The morphological language model is then interpolated with a modified Kneser-Ney trigram model. [sent-79, score-0.359]
45 In this interpolation the parameters λ depend on the cluster c2 of the his- tory word w2, P(w3|w1w2) i. [sent-80, score-0.092]
46 This setup may cause overfitting as every high frequent word w2 corresponds to a singleton class. [sent-83, score-0.134]
47 A grouping of several words into equivalence classes could therefore further improve the model; this, 526 however, is beyond the scope of this article. [sent-84, score-0.129]
48 4 Experimental Setup We compare the performance ofthe described model with a Kneser-Ney model and an interpolated model based on part-of-speech (POS) tags. [sent-87, score-0.208]
49 The relation between words and POS tags is many-to-many, but we transform it to a many-to-one relation by labeling every word independent of its context with its most frequent tag. [sent-88, score-0.073]
50 OOV words are treated equally even though their POS classes would not be known in a real application. [sent-89, score-0.129]
51 The experiments are carried out on a Wall Street Journal (WSJ) corpus of 50 million words that is split into training set (80%), valdev (5%), valtst (5%), and test set (10%). [sent-91, score-0.51]
52 The number of distinct feature vectors in training set, valdev and validation set (valdev+valtst) are 632, 466, and 512, respectively. [sent-92, score-0.318]
53 As mentioned above, the training set is used to learn suffixes and the maximum likelihood n-gram estimates. [sent-93, score-0.131]
54 The unknown word rate of the validation set is ǫ ≈ 0. [sent-94, score-0.212]
55 The first uses valdev for parameter estimation and valtst for testing and the second the entire validation set for parameter estimation and the test set for testing. [sent-97, score-0.535]
56 All models with a threshold greater or equal to the frequency of the most frequent word type are identical. [sent-98, score-0.139]
57 We use ∞ as the threshold to refer to these models. [sent-99, score-0.065]
58 IWn a ssiem ∞ilar a manner, tshheo lcdlu tsote rre count ∞ dee mnootdeesl a clustering wr mhearen ntwero, hweo crdlus are cino utnhet same ncoltuesste ar if and only if their features are identical. [sent-100, score-0.208]
59 This is the finest possible clustering of the feature vectors. [sent-101, score-0.216]
60 For small frequency thresholds overfitting effects cause that the interpolated models are worse than the KN model. [sent-105, score-0.137]
61 We can see that a clustering of the feature vectors is not necessary as the differences between all cluster models are small and c∞ is the overall best model. [sent-106, score-0.27]
62 Surprisingly, morphological clustering and POS classes are close even though over all events P(w3 |w1w2) of the valtst set. [sent-107, score-0.741]
63 On the right side, the subset of events where w1 or w2 are unknown is taken into account. [sent-108, score-0.153]
64 T|hwe overall best results for class models and POS models are highlighted in bold. [sent-109, score-0.037]
65 the POS class model uses oracle information to assign the right POS to an unknown word. [sent-110, score-0.24]
66 The optimal threshold is θ = 103 the bolded perplexity values 84. [sent-111, score-0.37]
67 35% of the word types were excluded from the morphological clustering (86% of the tokens). [sent-114, score-0.386]
68 – In a second evaluation we reduce the perplexity calculations to predictions of the form P(w3 |w1w2) where w1 or w2 are OOV words. [sent-116, score-0.253]
69 On such an event the KN model has to back off to a bigram or even unigram estimate, which results in inferior predictions and higher perplexity. [sent-117, score-0.102]
70 A first observation is that the perplexity of model c1 starts at a good value, but worsens with rising values for θ ≥ 10. [sent-120, score-0.369]
71 vTahleu reason iso tsehen sdwo mithin arniscine go fv proper nouns a 1n0d. [sent-121, score-0.047]
72 cardinal numbers at a frequency threshold ofone and in the distribution of OOV words (cf. [sent-122, score-0.279]
73 The c1 model with θ = 1 is specialized for predicting words after unknown nouns and cardinal numbers and two thirds of the unknown words are of exactly that type. [sent-124, score-0.613]
74 However, with rising θ, other word classes get a higher influence and different probability distributions are superimposed. [sent-125, score-0.161]
75 The best morphological model c∞ reduces the KN perplexity of 813. [sent-126, score-0.554]
76 In this case, we used the entire validation set for parameter tuning (i. [sent-130, score-0.059]
77 28, the perplexities for the best POS and clus527 ter model for θ = 1000 are 84. [sent-134, score-0.113]
78 For unknown histories the KN model perplexity is 767. [sent-137, score-0.535]
79 25 and the POS and c∞ cluster model perplexities at θ = 50 are 150. [sent-138, score-0.205]
80 Thus, the morphological model reduces perplexity by 81% compared to the KN model. [sent-141, score-0.554]
81 6 c∞ Conclusion We have presented a new class-based morphological language model. [sent-142, score-0.251]
82 In an experiment the model outperformed a modified Kneser-Ney model, especially in the prediction of the continuations of histories containing OOV words. [sent-143, score-0.129]
83 The model is entirely unsupervised, but works as well as a model using part-ofspeech information. [sent-144, score-0.1]
84 We plan to use our model for domain adaptation in applications like machine translation. [sent-146, score-0.05]
85 We then want to extend our model to other languages, which could be more challenging, as certain languages have a more complex morphology than English, but also worthwhile, if the unknown word rate is higher. [sent-147, score-0.291]
86 The model could be further improved by using contextual information for the word clustering and training a classifier based on morphological features to assign OOV words to these clusters. [sent-149, score-0.47]
87 Morph-based speech recognition and modeling of out-of-vocabulary words across languages. [sent-183, score-0.14]
88 An automatic technique to include grammatical and morphological information in a trigram-based statistical language model. [sent-191, score-0.251]
89 p g speech and automatically derived category-based language models for speech recognition. [sent-202, score-0.11]
90 Modeling morphologically rich languages using split words and unstructured dependencies. [sent-219, score-0.069]
wordName wordTfidf (topN-words)
[('oov', 0.459), ('kn', 0.288), ('valtst', 0.26), ('perplexity', 0.253), ('morphological', 0.251), ('valdev', 0.216), ('unknown', 0.153), ('clustering', 0.135), ('creutz', 0.132), ('suffixes', 0.131), ('pos', 0.108), ('capitalization', 0.097), ('classes', 0.095), ('cluster', 0.092), ('berton', 0.087), ('niesler', 0.087), ('steinbach', 0.087), ('histories', 0.079), ('bic', 0.076), ('factored', 0.074), ('bahl', 0.07), ('yuret', 0.07), ('souza', 0.07), ('cardinal', 0.07), ('vergyri', 0.07), ('morphemes', 0.07), ('shape', 0.07), ('rising', 0.066), ('threshold', 0.065), ('perplexities', 0.063), ('brown', 0.06), ('transition', 0.059), ('validation', 0.059), ('characters', 0.058), ('interpolated', 0.058), ('peter', 0.058), ('finnish', 0.058), ('kirchhoff', 0.058), ('bilmes', 0.056), ('clusters', 0.055), ('speech', 0.055), ('classbased', 0.054), ('morphology', 0.053), ('bolded', 0.052), ('inferior', 0.052), ('special', 0.052), ('singleton', 0.051), ('modeling', 0.051), ('model', 0.05), ('morpheme', 0.048), ('katrin', 0.048), ('nouns', 0.047), ('acoustics', 0.047), ('helmut', 0.046), ('schmid', 0.046), ('mercer', 0.046), ('emission', 0.045), ('overfitting', 0.044), ('feature', 0.043), ('cd', 0.041), ('german', 0.041), ('names', 0.04), ('frequent', 0.039), ('groups', 0.038), ('children', 0.038), ('hwe', 0.038), ('clar', 0.038), ('ofone', 0.038), ('inflecting', 0.038), ('ergun', 0.038), ('ici', 0.038), ('tshheo', 0.038), ('antti', 0.038), ('inw', 0.038), ('lalit', 0.038), ('dimitra', 0.038), ('finest', 0.038), ('hapax', 0.038), ('ller', 0.038), ('oovs', 0.038), ('pkn', 0.038), ('group', 0.037), ('numbers', 0.037), ('class', 0.037), ('arabic', 0.036), ('frequency', 0.035), ('ilar', 0.035), ('gart', 0.035), ('thirds', 0.035), ('pablo', 0.035), ('quency', 0.035), ('cino', 0.035), ('duce', 0.035), ('formerly', 0.035), ('iwn', 0.035), ('karypis', 0.035), ('languages', 0.035), ('words', 0.034), ('suffix', 0.034), ('signal', 0.033)]
simIndex simValue paperId paperTitle
same-paper 1 1.0000005 163 acl-2011-Improved Modeling of Out-Of-Vocabulary Words Using Morphological Classes
Author: Thomas Mueller ; Hinrich Schuetze
Abstract: We present a class-based language model that clusters rare words of similar morphology together. The model improves the prediction of words after histories containing outof-vocabulary words. The morphological features used are obtained without the use of labeled data. The perplexity improvement compared to a state of the art Kneser-Ney model is 4% overall and 81% on unknown histories.
2 0.27971935 203 acl-2011-Learning Sub-Word Units for Open Vocabulary Speech Recognition
Author: Carolina Parada ; Mark Dredze ; Abhinav Sethy ; Ariya Rastrow
Abstract: Large vocabulary speech recognition systems fail to recognize words beyond their vocabulary, many of which are information rich terms, like named entities or foreign words. Hybrid word/sub-word systems solve this problem by adding sub-word units to large vocabulary word based systems; new words can then be represented by combinations of subword units. Previous work heuristically created the sub-word lexicon from phonetic representations of text using simple statistics to select common phone sequences. We propose a probabilistic model to learn the subword lexicon optimized for a given task. We consider the task of out of vocabulary (OOV) word detection, which relies on output from a hybrid model. A hybrid model with our learned sub-word lexicon reduces error by 6.3% and 7.6% (absolute) at a 5% false alarm rate on an English Broadcast News and MIT Lectures task respectively.
3 0.25478411 175 acl-2011-Integrating history-length interpolation and classes in language modeling
Author: Hinrich Schutze
Abstract: Building on earlier work that integrates different factors in language modeling, we view (i) backing off to a shorter history and (ii) class-based generalization as two complementary mechanisms of using a larger equivalence class for prediction when the default equivalence class is too small for reliable estimation. This view entails that the classes in a language model should be learned from rare events only and should be preferably applied to rare events. We construct such a model and show that both training on rare events and preferable application to rare events improve perplexity when compared to a simple direct interpolation of class-based with standard language models.
4 0.20934308 208 acl-2011-Lexical Normalisation of Short Text Messages: Makn Sens a #twitter
Author: Bo Han ; Timothy Baldwin
Abstract: Twitter provides access to large volumes of data in real time, but is notoriously noisy, hampering its utility for NLP. In this paper, we target out-of-vocabulary words in short text messages and propose a method for identifying and normalising ill-formed words. Our method uses a classifier to detect ill-formed words, and generates correction candidates based on morphophonemic similarity. Both word similarity and context are then exploited to select the most probable correction candidate for the word. The proposed method doesn’t require any annotations, and achieves state-of-the-art performance over an SMS corpus and a novel dataset based on Twitter.
5 0.17732501 75 acl-2011-Combining Morpheme-based Machine Translation with Post-processing Morpheme Prediction
Author: Ann Clifton ; Anoop Sarkar
Abstract: This paper extends the training and tuning regime for phrase-based statistical machine translation to obtain fluent translations into morphologically complex languages (we build an English to Finnish translation system) . Our methods use unsupervised morphology induction. Unlike previous work we focus on morphologically productive phrase pairs – our decoder can combine morphemes across phrase boundaries. Morphemes in the target language may not have a corresponding morpheme or word in the source language. Therefore, we propose a novel combination of post-processing morphology prediction with morpheme-based translation. We show, using both automatic evaluation scores and linguistically motivated analyses of the output, that our methods outperform previously proposed ones and pro- vide the best known results on the EnglishFinnish Europarl translation task. Our methods are mostly language independent, so they should improve translation into other target languages with complex morphology. 1 Translation and Morphology Languages with rich morphological systems present significant hurdles for statistical machine translation (SMT) , most notably data sparsity, source-target asymmetry, and problems with automatic evaluation. In this work, we propose to address the problem of morphological complexity in an Englishto-Finnish MT task within a phrase-based translation framework. We focus on unsupervised segmentation methods to derive the morphological information supplied to the MT model in order to provide coverage on very large datasets and for languages with few hand-annotated 32 resources. In fact, in our experiments, unsupervised morphology always outperforms the use of a hand-built morphological analyzer. Rather than focusing on a few linguistically motivated aspects of Finnish morphological behaviour, we develop techniques for handling morphological complexity in general. We chose Finnish as our target language for this work, because it exemplifies many of the problems morphologically complex languages present for SMT. Among all the languages in the Europarl data-set, Finnish is the most difficult language to translate from and into, as was demonstrated in the MT Summit shared task (Koehn, 2005) . Another reason is the current lack of knowledge about how to apply SMT successfully to agglutinative languages like Turkish or Finnish. Our main contributions are: 1) the introduction of the notion of segmented translation where we explicitly allow phrase pairs that can end with a dangling morpheme, which can connect with other morphemes as part of the translation process, and 2) the use of a fully segmented translation model in combination with a post-processing morpheme prediction system, using unsupervised morphology induction. Both of these approaches beat the state of the art on the English-Finnish translation task. Morphology can express both content and function categories, and our experiments show that it is important to use morphology both within the translation model (for morphology with content) and outside it (for morphology contributing to fluency) . Automatic evaluation measures for MT, BLEU (Papineni et al., 2002), WER (Word Error Rate) and PER (Position Independent Word Error Rate) use the word as the basic unit rather than morphemes. In a word comProce dPinogrstla ofn tdh,e O 4r9etghon A,n Jnu nael 1 M9-e 2t4i,n2g 0 o1f1 t.he ?c A2s0s1o1ci Aatsiosonc fioartio Cno fmorpu Ctoamtiopnuatalt Lioin gauli Lsitnicgsu,i psatgices 32–42, prised of multiple morphemes, getting even a single morpheme wrong means the entire word is wrong. In addition to standard MT evaluation measures, we perform a detailed linguistic analysis of the output. Our proposed approaches are significantly better than the state of the art, achieving the highest reported BLEU scores on the English-Finnish Europarl version 3 data-set. Our linguistic analysis shows that our models have fewer morpho-syntactic errors compared to the word-based baseline. 2 2.1 Models Baseline Models We set up three baseline models for comparison in this work. The first is a basic wordbased model (called Baseline in the results) ; we trained this on the original unsegmented version of the text. Our second baseline is a factored translation model (Koehn and Hoang, 2007) (called Factored) , which used as factors the word, “stem” 1 and suffix. These are derived from the same unsupervised segmentation model used in other experiments. The results (Table 3) show that a factored model was unable to match the scores of a simple wordbased baseline. We hypothesize that this may be an inherently difficult representational form for a language with the degree of morphological complexity found in Finnish. Because the morphology generation must be precomputed, for languages with a high degree of morphological complexity, the combinatorial explosion makes it unmanageable to capture the full range of morphological productivity. In addition, because the morphological variants are generated on a per-word basis within a given phrase, it excludes productive morphological combination across phrase boundaries and makes it impossible for the model to take into account any longdistance dependencies between morphemes. We conclude from this result that it may be more useful for an agglutinative language to use morphology beyond the confines of the phrasal unit, and condition its generation on more than just the local target stem. In order to compare the 1see Section 2.2. 33 performance of unsupervised segmentation for translation, our third baseline is a segmented translation model based on a supervised segmentation model (called Sup) , using the hand-built Omorfi morphological analyzer (Pirinen and Listenmaa, 2007) , which provided slightly higher BLEU scores than the word-based baseline. 2.2 Segmented Translation For segmented translation models, it cannot be taken for granted that greater linguistic accuracy in segmentation yields improved translation (Chang et al. , 2008) . Rather, the goal in segmentation for translation is instead to maximize the amount of lexical content-carrying morphology, while generalizing over the information not helpful for improving the translation model. We therefore trained several different segmentation models, considering factors of granularity, coverage, and source-target symmetry. We performed unsupervised segmentation of the target data, using Morfessor (Creutz and Lagus, 2005) and Paramor (Monson, 2008) , two top systems from the Morpho Challenge 2008 (their combined output was the Morpho Challenge winner) . However, translation models based upon either Paramor alone or the combined systems output could not match the wordbased baseline, so we concentrated on Morfessor. Morfessor uses minimum description length criteria to train a HMM-based segmentation model. When tested against a human-annotated gold standard of linguistic morpheme segmentations for Finnish, this algorithm outperforms competing unsupervised methods, achieving an F-score of 67.0% on a 3 million sentence corpus (Creutz and Lagus, 2006) . Varying the perplexity threshold in Morfessor does not segment more word types, but rather over-segments the same word types. In order to get robust, common segmentations, we trained the segmenter on the 5000 most frequent words2 ; we then used this to segment the entire data set. In order to improve coverage, we then further segmented 2For the factored model baseline we also used the same setting perplexity = 30, 5,000 most frequent words, but with all but the last suffix collapsed and called the “stem” . TabHMleoat1nr:gplhiMngor phermphocTur631ae04in, 81c9ie03ns67gi,64n0S14e567theTp 2rsa51t, 29Se 3t168able and in translation. any word type that contained a match from the most frequent suffix set, looking for the longest matching suffix character string. We call this method Unsup L-match. After the segmentation, word-internal morpheme boundary markers were inserted into the segmented text to be used to reconstruct the surface forms in the MT output. We then trained the Moses phrase-based system (Koehn et al., 2007) on the segmented and marked text. After decoding, it was a simple matter to join together all adjacent morphemes with word-internal boundary markers to reconstruct the surface forms. Figure 1(a) gives the full model overview for all the variants of the segmented translation model (supervised/unsupervised; with and without the Unsup L-match procedure) . Table 1shows how morphemes are being used in the MT system. Of the phrases that included segmentations (‘Morph’ in Table 1) , roughly a third were ‘productive’, i.e. had a hanging morpheme (with a form such as stem+) that could be joined to a suffix (‘Hanging Morph’ in Table 1) . However, in phrases used while decoding the development and test data, roughly a quarter of the phrases that generated the translated output included segmentations, but of these, only a small fraction (6%) had a hanging morpheme; and while there are many possible reasons to account for this we were unable to find a single convincing cause. 2.3 Morphology Generation Morphology generation as a post-processing step allows major vocabulary reduction in the translation model, and allows the use of morphologically targeted features for modeling inflection. A possible disadvantage of this approach is that in this model there is no opportunity to con34 sider the morphology in translation since it is removed prior to training the translation model. Morphology generation models can use a variety of bilingual and contextual information to capture dependencies between morphemes, often more long-distance than what is possible using n-gram language models over morphemes in the segmented model. Similar to previous work (Minkov et al. , 2007; Toutanova et al. , 2008) , we model morphology generation as a sequence learning problem. Un- like previous work, we use unsupervised morphology induction and use automatically generated suffix classes as tags. The first phase of our morphology prediction model is to train a MT system that produces morphologically simplified word forms in the target language. The output word forms are complex stems (a stem and some suffixes) but still missing some important suffix morphemes. In the second phase, the output of the MT decoder is then tagged with a sequence of abstract suffix tags. In particular, the output of the MT decoder is a sequence of complex stems denoted by x and the output is a sequence of suffix class tags denoted by y. We use a list of parts from (x,y) and map to a d-dimensional feature vector Φ(x, y) , with each dimension being a real number. We infer the best sequence of tags using: F(x) = argymaxp(y | x,w) where F(x) returns the highest scoring output y∗ . A conditional random field (CRF) (Lafferty et al. , 2001) defines the conditional probability as a linear score for each candidate y and a global normalization term: logp(y | x, w) = Φ(x, y) · w − log Z where Z = Py0∈ exp(Φ(x, y0) · w) . We use stochastiPc gradient descent (using crfsgd3) to train the weight vector w. So far, this is all off-the-shelf sequence learning. However, the output y∗ from the CRF decoder is still only a sequence of abstract suffix tags. The third and final phase in our morphology prediction model GEN(x) 3 http://leon. bottou. org/projects/sgd English Training Data words Finnish Training Data words Morphological Pre-Processing stem+ +morph MT System Alignment: word word word stem+ +morph stem stem+ +morph Post-Process: Morph Re-Stitching Fully inflected surface form Evaluation against original reference (a) Segmented Translation Model English Training Data words Finnish Training Data Morphological Pre-Prowceosrdsisng 1 stem+ +morph1+ +morph2 Morphological Pre-Processing 2 stem+ +morph1+ MPosrpthe-mPRr+eo-+cSmetsio crhp1i:nhg+swteomrd+ MA+lTmigwnSomyrspdthen 1mt:+ wsotermd complex stem: stem+morph1+ MPo rpsht-oPlro gcyesGse2n:erCaRtioFnstem+morph1+ morph2sLuarnfagcueagfeorMmomdealp ing Fully inflected surface form Evaluation against original reference (b) Post-Processing Model Translation & Generation Figure 1: Training and testing pipelines for the SMT models. is to take the abstract suffix tag sequence y∗ and then map it into fully inflected word forms, and rank those outputs using a morphemic language model. The abstract suffix tags are extracted from the unsupervised morpheme learning process, and are carefully designed to enable CRF training and decoding. We call this model CRFLM for short. Figure 1(b) shows the full pipeline and Figure 2 shows a worked example of all the steps involved. We use the morphologically segmented training data (obtained using the segmented corpus described in Section 2.24) and remove selected suffixes to create a morphologically simplified version of the training data. The MT model is trained on the morphologically simplified training data. The output from the MT system is then used as input to the CRF model. The CRF model was trained on a ∼210,000 Finnish sentences, consisting noefd d∼ o1n.5 a am ∼il2li1o0n,0 tokens; tishhe 2,000 cseens,te cnoncse Europarl t.e5s tm isl eito nco tnoskiesntesd; hoef 41,434 stem tokens. The labels in the output sequence y were obtained by selecting the most productive 150 stems, and then collapsing certain vowels into equivalence classes corresponding to Finnish vowel harmony patterns. Thus 4Note that unlike Section 2.2 we do not use Unsup L-match because when evaluating the CRF model on the suffix prediction task it obtained 95.61% without using Unsup L-match and 82.99% when using Unsup L-match. 35 variants -k¨ o and -ko become vowel-generic enclitic particle -kO, and variants -ss ¨a and -ssa become the vowel-generic inessive case marker -ssA, etc. This is the only language-specific component of our translation model. However, we expect this approach to work for other agglutinative languages as well. For fusional languages like Spanish, another mapping from suffix to abstract tags might be needed. These suffix transformations to their equivalence classes prevent morphophonemic variants of the same morpheme from competing against each other in the prediction model. This resulted in 44 possible label outputs per stem which was a reasonable sized tag-set for CRF training. The CRF was trained on monolingual features of the segmented text for suffix prediction, where t is the current token: Word Stem st−n, .., st, .., st+n(n = 4) Morph Prediction yt−2 , yt−1 , yt With this simple feature set, we were able to use features over longer distances, resulting in a total of 1,110,075 model features. After CRF based recovery of the suffix tag sequence, we use a bigram language model trained on a full segmented version on the training data to recover the original vowels. We used bigrams only, because the suffix vowel harmony alternation depends only upon the preceding phonemes in the word from which it was segmented. original training koskevaa mietint o¨ ¨a data: k ¨asitell ¨a ¨an segmentation: koske+ +va+ +a mietint ¨o+ + a¨ k a¨si+ +te+ +ll a¨+ + a¨+ +n (train bigram language model with mapping A = { a , a }) map n fi bniaglr asmuff liaxn gtou agbest mraocdte tag-set: koske+ +va+ +A mietint ¨o+ +A k ¨asi+ +te+ +ll ¨a+ + ¨a+ +n (train CRF model to predict the final suffix) peeling of final suffix: koske+ +va+ mietint ¨o+ k a¨si+ +te+ +ll a¨+ + a¨+ (train SMT model on this transformation of training data) (a) Training decoder output: koske+ +va+ mietint o¨+ k a¨si+ +te+ +ll a¨+ + a¨+ decoder output stitched up: koskeva+ mietint o¨+ k ¨asitell ¨a ¨a+ CRF model prediction: x = ‘koskeva+ mietint ¨o+ k ¨asitell ¨a ¨a+’, y = ‘+A +A +n’ koskeva+ +A mietint ¨o+ +A k ¨asitell a¨ ¨a+ +n unstitch morphemes: koske+ +va+ +A mietint ¨o+ +A k ¨asi+ +te+ +ll ¨a+ + ¨a+ +n language model disambiguation: koske+ +va+ +a mietint ¨o+ + a¨ k a¨si+ +te+ +ll a¨+ + a¨+ +n final stitching: koskevaa mietint o¨ ¨a k ¨asitell ¨a ¨an (the output is then compared to the reference translation) (b) Decoding Figure 2: Worked example of all steps in the post-processing morphology prediction model. 3 Experimental Results used the Europarl version 3 corpus (Koehn, 2005) English-Finnish training data set, as well as the standard development and test data sets. Our parallel training data consists of ∼1 million senFor all of the models built in this paper, we tpeanrcaelsle lo tfr a4i0n nwgor ddast or less, sw ohfi ∼le 1t mhei development and test sets were each 2,000 sentences long. In all the experiments conducted in this paper, we used the Moses5 phrase-based translation system (Koehn et al. , 2007) , 2008 version. We trained all of the Moses systems herein using the standard features: language model, reordering model, translation model, and word penalty; in addition to these, the factored experiments called for additional translation and generation features for the added factors as noted above. We used in all experiments the following settings: a hypothesis stack size 100, distortion limit 6, phrase translations limit 20, and maximum phrase length 20. For the language models, we used SRILM 5-gram language models (Stolcke, 2002) for all factors. For our word-based Baseline system, we trained a word-based model using the same Moses system with identical settings. For evaluation against segmented translation systems in segmented forms before word reconstruction, we also segmented the baseline system’s word-based output. All the BLEU scores reported are for lowercase evaluation. We did an initial evaluation of the segmented output translation for each system using the no5http://www.statmt.org/moses/ 36 TabSlBUeuna2gps:meulSipengLmta-e nioatedchMo12dme804-.lB8S714cL±oEr0eUs.6 9 S8up19Nre.358ofe498rUs9ntoihe supervised segmentation baseline model. m-BLEU indicates that the segmented output was evaluated against a segmented version of the reference (this measure does not have the same correlation with human judgement as BLEU) . No Uni indicates the segmented BLEU score without unigrams. tion of m-BLEU score (Luong et al. , 2010) where the BLEU score is computed by comparing the segmented output with a segmented reference translation. Table 2 shows the m-BLEU scores for various systems. We also show the m-BLEU score without unigrams, since over-segmentation could lead to artificially high m-BLEU scores. In fact, if we compare the relative improvement of our m-BLEU scores for the Unsup L-match system we see a relative improvement of 39.75% over the baseline. Luong et. al. (2010) report an m-BLEU score of 55.64% but obtain a relative improvement of 0.6% over their baseline m-BLEU score. We find that when using a good segmentation model, segmentation of the morphologically complex target language improves model performance over an unsegmented baseline (the confidence scores come from bootstrap resampling) . Table 3 shows the evaluation scores for all the baselines and the methods introduced in this paper using standard wordbased lowercase BLEU, WER and PER. We do TSCMaFBU(LubanRolpcesdFotu3lne-ipLr:gMdeLT-tms.al,Stc2ho0r1es:)l 1wB54 Le.r682E90c 27a9Us∗eBL-7 W46E3. U659478R6,1WE-7 TR412E. 847Ra1528nd TER. The ∗ indicates a statistically significant improvement o∗f BndLiEcaUte score over tchalel yB saisgenli nfice mntod imel.The boldface scores are the best performing scores per evaluation measure. better than (Luong et al. , 2010) , the previous best score for this task. We also show a better relative improvement over our baseline when compared to (Luong et al., 2010) : a relative improvement of 4.86% for Unsup L-match compared to our baseline word-based model, compared to their 1.65% improvement over their baseline word-based model. Our best performing method used unsupervised morphology with L-match (see Section 2.2) and the improvement is significant: bootstrap resampling provides a confidence margin of ±0.77 and a t-test (Collins ceot nafli.d , 2005) sahrogwined o significance aw ti-thte p = 0o.0ll0in1s. 3.1 Morphological Fluency Analysis To see how well the models were doing at getting morphology right, we examined several patterns of morphological behavior. While we wish to explore minimally supervised morphological MT models, and use as little language specific information as possible, we do want to use linguistic analysis on the output of our system to see how well the models capture essential morphological information in the target language. So, we ran the word-based baseline system, the segmented model (Unsup L-match) , and the prediction model (CRF-LM) outputs, along with the reference translation through the supervised morphological analyzer Omorfi (Pirinen and Listenmaa, 2007) . Using this analysis, we looked at a variety of linguistic constructions that might reveal patterns in morphological behavior. These were: (a) explicitly marked 37 noun forms, (b) noun-adjective case agreement, (c) subject-verb person/number agreement, (d) transitive object case marking, (e) postpositions, and (f) possession. In each of these categories, we looked for construction matches on a per-sentence level between the models’ output and the reference translation. Table 4 shows the models’ performance on the constructions we examined. In all of the categories, the CRF-LM model achieves the best precision score, as we explain below, while the Unsup L-match model most frequently gets the highest recall score. A general pattern in the most prevalent of these constructions is that the baseline tends to prefer the least marked form for noun cases (corresponding to the nominative) more than the reference or the CRF-LM model. The baseline leaves nouns in the (unmarked) nominative far more than the reference, while the CRF-LM model comes much closer, so it seems to fare better at explicitly marking forms, rather than defaulting to the more frequent unmarked form. Finnish adjectives must be marked with the same case as their head noun, while verbs must agree in person and number with their subject. We saw that in both these categories, the CRFLM model outperforms for precision, while the segmented model gets the best recall. In addition, Finnish generally marks direct objects of verbs with the accusative or the partitive case; we observed more accusative/partitive-marked nouns following verbs in the CRF-LM output than in the baseline, as illustrated by example (1) in Fig. 3. While neither translation picks the same verb as in the reference for the input ‘clarify,’ the CRFLM-output paraphrases it by using a grammatical construction of the transitive verb followed by a noun phrase inflected with the accusative case, correctly capturing the transitive construction. The baseline translation instead follows ‘give’ with a direct object in the nominative case. To help clarify the constructions in question, we have used Google Translate6 to provide back6 http://translate.google. com/ of occurrences per sentence, recall and F-score. also averaged The constructions over the various translations. are listed in descending P, R and F stand for precision, order of their frequency in the texts. The highlighted value in each column is the most accurate with respect to the reference value. translations of our MT output into English; to contextualize these back-translations, we have provided Google’s back-translation of the reference. The use of postpositions shows another difference between the models. Finnish postpositions require the preceding noun to be in the genitive or sometimes partitive case, which occurs correctly more frequently in the CRF-LM than the baseline. In example (2) in Fig. 3, all three translations correspond to the English text, ‘with the basque nationalists. ’ However, the CRF-LM output is more grammatical than the baseline, because not only do the adjective and noun agree for case, but the noun ‘baskien’ to which the postposition ‘kanssa’ belongs is marked with the correct genitive case. However, this well-formedness is not rewarded by BLEU, because ‘baskien’ does not match the reference. In addition, while Finnish may express possession using case marking alone, it has another construction for possession; this can disambiguate an otherwise ambiguous clause. This alternate construction uses a pronoun in the genitive case followed by a possessive-marked noun; we see that the CRF-LM model correctly marks this construction more frequently than the baseline. As example (3) in Fig. 3 shows, while neither model correctly translates ‘matkan’ (‘trip’) , the baseline’s output attributes the inessive ‘yhteydess’ (‘connection’) as belonging to ‘tulokset’ (‘results’) , and misses marking the possession linking it to ‘Commissioner Fischler’. Our manual evaluation shows that the CRF38 LM model is producing output translations that are more morphologically fluent than the wordbased baseline and the segmented translation Unsup L-match system, even though the word choices lead to a lower BLEU score overall when compared to Unsup L-match. 4 Related Work The work on morphology in MT can be grouped into three categories, factored models, segmented translation, and morphology generation. Factored models (Koehn and Hoang, 2007) factor the phrase translation probabilities over additional information annotated to each word, allowing for text to be represented on multiple levels of analysis. We discussed the drawbacks of factored models for our task in Section 2. 1. While (Koehn and Hoang, 2007; Yang and Kirchhoff, 2006; Avramidis and Koehn, 2008) obtain improvements using factored models for translation into English, German, Spanish, and Czech, these models may be less useful for capturing long-distance dependencies in languages with much more complex morphological systems such as Finnish. In our experiments factored models did worse than the baseline. Segmented translation performs morphological analysis on the morphologically complex text for use in the translation model (Brown et al. , 1993; Goldwater and McClosky, 2005; de Gispert and Mari n˜o, 2008) . This method unpacks complex forms into simpler, more frequently occurring components, and may also increase the symmetry of the lexically realized content be(1) Input: ‘the charter we are to approve today both strengthens and gives visible shape to the common fundamental rights and values our community is to be based upon. ’ a. Reference: perusoikeuskirja , jonka t ¨an ¨a ¨an aiomme hyv a¨ksy ¨a , sek ¨a vahvistaa ett ¨a selvent a¨ a¨ (selvent ¨a a¨/VERB/ACT/INF/SG/LAT-clarify) niit a¨ (ne/PRONOUN/PL/PAR-them) yhteisi ¨a perusoikeuksia ja arvoja , joiden on oltava yhteis¨ omme perusta. Back-translation: ‘Charter of Fundamental Rights, which today we are going to accept that clarify and strengthen the common fundamental rights and values, which must be community based. ’ b. Baseline: perusoikeuskirja me hyv ¨aksymme t¨ an ¨a a¨n molemmat vahvistaa ja antaa (antaa/VERB/INF/SG/LATgive) n a¨kyv a¨ (n¨ aky a¨/VERB/ACT/PCP/SG/NOM-visible) muokata yhteist ¨a perusoikeuksia ja arvoja on perustuttava. Back-translation: ‘Charter today, we accept both confirm and modify to make a visible and common values, fundamental rights must be based. ’ c. CRF-LM: perusoikeuskirja on hyv a¨ksytty t ¨an ¨a ¨an , sek ¨a vahvistaa ja antaa (antaa/VERB/ACT/INF/SG/LAT-give) konkreettisen (konkreettinen/ADJECTIVE/SG/GEN,ACC-concrete) muodon (muoto/NOUN/SG/GEN,ACCshape) yhteisi ¨a perusoikeuksia ja perusarvoja , yhteis¨ on on perustuttava. Back-translation: ‘Charter has been approved today, and to strengthen and give concrete shape to the common basic rights and fundamental values, the Community must be based. ’ (2) Input: ‘with the basque nationalists’ a. Reference: baskimaan kansallismielisten kanssa basque-SG/NOM+land-SG/GEN,ACC nationalists-PL/GEN with-POST b. Baseline: baskimaan kansallismieliset kanssa basque-SG/NOM-+land-SG/GEN,ACC kansallismielinen-PL/NOM,ACC-nationalists POST-with c. CRF-LM: kansallismielisten baskien kanssa nationalists-PL/GEN basques-PL/GEN with-POST (3) Input: ‘and in this respect we should value the latest measures from commissioner fischler , the results of his trip to morocco on the 26th of last month and the high level meetings that took place, including the one with the king himself’ a. Reference: ja t ¨ass¨ a mieless ¨a osaamme my¨ os arvostaa komission j¨ asen fischlerin viimeisimpi ¨a toimia , jotka ovat h a¨nen (h¨ anen/GEN-his) marokkoon 26 lokakuuta tekemns (tekem¨ ans ¨a/POSS-his) matkan (matkan/GENtour) ja korkean tason kokousten jopa itsens¨ a kuninkaan kanssa tulosta Back-translation: ‘and in this sense we can also appreciate the Commissioner Fischler’s latest actions, which are his to Morocco 26 October trip to high-level meetings and even the king himself with the result b. Baseline: ja t ¨ass¨ a yhteydess a¨ olisi arvoa viimeisin toimia komission j¨ asen fischler , tulokset monitulkintaisia marokon yhteydess a¨ (yhteydess/INE-connection) , ja viime kuussa pidettiin korkean tason kokouksissa , mukaan luettuna kuninkaan kanssa Back-translation: ‘and in this context would be the value of the last act, Commissioner Fischler, the results of the Moroccan context, ambiguous, and last month held high level meetings, including with the king’ c. CRF-LM: ja t ¨ass¨ a yhteydess a¨ meid ¨an olisi lis ¨aarvoa viimeist ¨a toimenpiteit a¨ kuin komission j¨ asen fischler , ett a¨ h a¨nen (h¨ anen/GEN-his) kokemuksensa (kokemuksensa/POSS-experience) marokolle (marokolle-Moroccan) viime kuun 26 ja korkean tason tapaamiset j¨ arjestettiin, kuninkaan kanssa Back-translation: ‘and in this context, we should value the last measures as the Commissioner Fischler, that his experience in Morocco has on the 26th and high-level meetings took place, including with the king. ’ Figure 3: Morphological fluency analysis (see Section 3. 1) . tween source and target. In a somewhat orthogonal approach to ours, (Ma et al. , 2007) use alignment of a parallel text to pack together adjacent segments in the alignment output, which are then fed back to the word aligner to bootstrap an improved alignment, which is then used in the translation model. We compared our results against (Luong et al. , 2010) in Table 3 since their results are directly comparable to ours. They use a segmented phrase table and language model along with the word-based versions in the decoder and in tuning a Finnish target. Their approach requires segmented phrases 39 to match word boundaries, eliminating morphologically productive phrases. In their work a segmented language model can score a translation, but cannot insert morphology that does not show source-side reflexes. In order to perform a similar experiment that still allowed for morphologically productive phrases, we tried training a segmented translation model, the output of which we stitched up in tuning so as to tune to a word-based reference. The goal of this experiment was to control the segmented model’s tendency to overfit by rewarding it for using correct whole-word forms. However, we found that this approach was less successful than using the segmented reference in tuning, and could not meet the baseline (13.97% BLEU best tuning score, versus 14.93% BLEU for the baseline best tuning score) . Previous work in segmented translation has often used linguistically motivated morphological analysis selectively applied based on a language-specific heuristic. A typical approach is to select a highly inflecting class of words and segment them for particular morphology (de Gispert and Mari n˜o, 2008; Ramanathan et al. , 2009) . Popovi¸ c and Ney (2004) perform segmentation to reduce morphological complexity of the source to translate into an isolating target, reducing the translation error rate for the English target. For Czech-to-English, Goldwater and McClosky (2005) lemmatized the source text and inserted a set of ‘pseudowords’ expected to have lexical reflexes in English. Minkov et. al. (2007) and Toutanova et. al. (2008) use a Maximum Entropy Markov Model for morphology generation. The main drawback to this approach is that it removes morphological information from the translation model (which only uses stems) ; this can be a problem for languages in which morphology ex- presses lexical content. de Gispert (2008) uses a language-specific targeted morphological classifier for Spanish verbs to avoid this issue. Talbot and Osborne (2006) use clustering to group morphological variants of words for word alignments and for smoothing phrase translation tables. Habash (2007) provides various methods to incorporate morphological variants of words in the phrase table in order to help recognize out of vocabulary words in the source language. 5 Conclusion and Future Work We found that using a segmented translation model based on unsupervised morphology induction and a model that combined morpheme segments in the translation model with a postprocessing morphology prediction model gave us better BLEU scores than a word-based baseline. Using our proposed approach we obtain better scores than the state of the art on the EnglishFinnish translation task (Luong et al. , 2010) : from 14.82% BLEU to 15.09%, while using a 40 simpler model. We show that using morphological segmentation in the translation model can improve output translation scores. We also demonstrate that for Finnish (and possibly other agglutinative languages) , phrase-based MT benefits from allowing the translation model access to morphological segmentation yielding productive morphological phrases. Taking advantage of linguistic analysis of the output we show that using a post-processing morphology generation model can improve translation fluency on a sub-word level, in a manner that is not captured by the BLEU word-based evaluation measure. In order to help with replication of the results in this paper, we have run the various morphological analysis steps and created the necessary training, tuning and test data files needed in order to train, tune and test any phrase-based machine translation system with our data. The files can be downloaded from natlang. cs.sfu. ca. In future work we hope to explore the utility of phrases with productive morpheme boundaries and explore why they are not used more pervasively in the decoder. Evaluation measures for morphologically complex languages and tun- ing to those measures are also important future work directions. Also, we would like to explore a non-pipelined approach to morphological preand post-processing so that a globally trained model could be used to remove the target side morphemes that would improve the translation model and then predict those morphemes in the target language. Acknowledgements This research was partially supported by NSERC, Canada (RGPIN: 264905) and a Google Faculty Award. We would like to thank Christian Monson, Franz Och, Fred Popowich, Howard Johnson, Majid Razmara, Baskaran Sankaran and the anonymous reviewers for their valuable comments on this work. We would particularly like to thank the developers of the open-source Moses machine translation toolkit and the Omorfi morphological analyzer for Finnish which we used for our experiments. References Eleftherios Avramidis and Philipp Koehn. 2008. Enriching morphologically poor languages for statistical machine translation. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, page 763?770, Columbus, Ohio, USA. Association for Computational Linguistics. Peter F. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, and R. L. Mercer. 1993. The mathematics of statistical machine translation: Parameter estimation. Computational Linguistics, 19(2) :263–31 1. Pi-Chuan Chang, Michel Galley, and Christopher D. Manning. 2008. Optimizing Chinese word segmentation for machine translation performance. In Proceedings of the Third Workshop on Statistical Machine Translation, pages 224–232, Columbus, Ohio, June. Association for Computational Linguistics. Michael Collins, Philipp Koehn, and Ivona Kucerova. 2005. Clause restructuring for statistical machine translation. In Proceedings of 43rd Annual Meeting of the Association for Computational Linguistics (A CL05). Association for Computational Linguistics. Mathias Creutz and Krista Lagus. 2005. Inducing the morphological lexicon of a natural language from unannotated text. In Proceedings of the International and Interdisciplinary Conference on Adaptive Knowledge Representation and Reason- ing (AKRR ’05), pages 106–113, Espoo, Finland. Mathias Creutz and Krista Lagus. 2006. Morfessor in the morpho challenge. In Proceedings of the PASCAL Challenge Workshop on Unsupervised Segmentation of Words into Morphemes. Adri ´a de Gispert and Jos e´ Mari n˜o. 2008. On the impact of morphology in English to Spanish statistical MT. Speech Communication, 50(11-12) . Sharon Goldwater and David McClosky. 2005. Improving statistical MT through morphological analysis. In Proceedings of the Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing, pages 676–683, Vancouver, B.C. , Canada. Association for Computational Linguistics. Philipp Koehn and Hieu Hoang. 2007. Factored translation models. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 868–876, Prague, Czech Republic. Association for Computational Linguistics. 41 Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, Chris Dyer, Ondrej Bojar, Alexandra Constantin, and Evan Herbst. 2007. Moses: Open source toolkit for statistical machine translation. In A CL ‘07: Proceedings of the 45th Annual Meeting of the A CL on Interactive Poster and Demonstration Sessions, pages 177–108, Prague, Czech Republic. Association for Computational Linguistics. Philipp Koehn. 2005. Europarl: A parallel corpus for statistical machine translation. In Proceedings of Machine Translation Summit X, pages 79–86, Phuket, Thailand. Association for Computational Linguistics. John Lafferty, Andrew McCallum, and Fernando Pereira. 2001. Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In Proceedings of the 18th International Conference on Machine Learning, pages 282–289, San Francisco, California, USA. Association for Computing Machinery. Minh-Thang Luong, Preslav Nakov, and Min-Yen Kan. 2010. A hybrid morpheme-word representation for machine translation of morphologically rich languages. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 148–157, Cambridge, Massachusetts. Association for Computational Linguistics. Yanjun Ma, Nicolas Stroppa, and Andy Way. 2007. Bootstrapping word alignment via word packing. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 304–311, Prague, Czech Republic. Association for Computational Linguistics. Einat Minkov, Kristina Toutanova, and Hisami Suzuki. 2007. Generating complex morphology for machine translation. In In Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics (A CL07), pages 128–135, Prague, Czech Republic. Association for Computational Linguistics. Christian Monson. 2008. Paramor and morpho challenge 2008. In Lecture Notes in Computer Science: Workshop of the Cross-Language Evaluation Forum (CLEF 2008), Revised Selected Papers. Habash Nizar. 2007. Four techniques for online handling of out-of-vocabulary words in arabic-english statistical machine translation. In Proceedings of the 46th Annual Meeting of the Association of Computational Linguistics, Columbus, Ohio. Association for Computational Linguistics. Kishore Papineni, Salim Roukos, Todd Ward, and Zhu. 2002. BLEU: A method for automatic evaluation of machine translation. In Proceedings of 40th Annual Meeting of the Association for Computational Linguistics A CL, pages 311–318, Philadelphia, Pennsylvania, USA. Association for Computational Linguistics. Tommi Pirinen and Inari Listenmaa. 2007. Omorfi morphological analzer. http://gna.org/projects/omorfi. Maja Popovi¸ c and Hermann Ney. 2004. Towards the use of word stems and suffixes for statistiWei jing cal machine translation. In Proceedings of the 4th International Conference on Language Resources and Evaluation (LREC), pages 1585–1588, Lisbon, Portugal. European Language Resources Association (ELRA) . Ananthakrishnan Ramanathan, Hansraj Choudhary, Avishek Ghosh, and Pushpak Bhattacharyya. 2009. Case markers and morphology: Addressing the crux of the fluency problem in EnglishHindi SMT. In Proceedings of the Joint Conference of the 4 7th Annual Meeting of the Association for Computational Linguistics and the 4th International Joint Conference on Natural Language Processing of the Asian Federation of Natural Language Processing, pages 800–808, Suntec, Singapore. Association for Computational Linguistics. Andreas Stolcke. 2002. Srilm – an extensible language modeling toolkit. 7th International Conference on Spoken Language Processing, 3:901–904. David Talbot and Miles Osborne. 2006. Modelling lexical redundancy for machine translation. In Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics, pages 969–976, Sydney, Australia, July. Association for Computational Linguistics. Kristina Toutanova, Hisami Suzuki, and Achim Ruopp. 2008. Applying morphology generation models to machine translation. In Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 514–522, Columbus, Ohio, USA. Association for Computational Linguistics. Mei Yang and Katrin Kirchhoff. 2006. Phrase-based backoff models for machine translation of highly inflected languages. In Proceedings of the European Chapter of the Association for Computational Linguistics, pages 41–48, Trento, Italy. Association for Computational Linguistics. 42
6 0.17115535 10 acl-2011-A Discriminative Model for Joint Morphological Disambiguation and Dependency Parsing
7 0.15272586 318 acl-2011-Unsupervised Bilingual Morpheme Segmentation and Alignment with Context-rich Hidden Semi-Markov Models
8 0.12311565 24 acl-2011-A Scalable Probabilistic Classifier for Language Modeling
9 0.11542859 104 acl-2011-Domain Adaptation for Machine Translation by Mining Unseen Words
10 0.1048243 124 acl-2011-Exploiting Morphology in Turkish Named Entity Recognition System
11 0.095916897 164 acl-2011-Improving Arabic Dependency Parsing with Form-based and Functional Morphological Features
12 0.09477818 29 acl-2011-A Word-Class Approach to Labeling PSCFG Rules for Machine Translation
13 0.093214348 142 acl-2011-Generalized Interpolation in Decision Tree LM
14 0.092377201 233 acl-2011-On-line Language Model Biasing for Statistical Machine Translation
15 0.08610896 193 acl-2011-Language-independent compound splitting with morphological operations
16 0.084598221 14 acl-2011-A Hierarchical Model of Web Summaries
17 0.082838424 39 acl-2011-An Ensemble Model that Combines Syntactic and Semantic Clustering for Discriminative Dependency Parsing
18 0.082059018 238 acl-2011-P11-2093 k2opt.pdf
19 0.081115089 44 acl-2011-An exponential translation model for target language morphology
20 0.078246325 329 acl-2011-Using Deep Morphology to Improve Automatic Error Detection in Arabic Handwriting Recognition
topicId topicWeight
[(0, 0.195), (1, -0.015), (2, -0.039), (3, -0.02), (4, -0.033), (5, -0.01), (6, 0.139), (7, -0.008), (8, 0.049), (9, 0.237), (10, -0.069), (11, 0.051), (12, -0.071), (13, 0.114), (14, 0.105), (15, -0.147), (16, -0.169), (17, 0.07), (18, 0.084), (19, 0.062), (20, 0.114), (21, -0.158), (22, 0.065), (23, -0.197), (24, 0.022), (25, -0.07), (26, 0.081), (27, 0.005), (28, -0.107), (29, 0.08), (30, -0.017), (31, -0.229), (32, -0.082), (33, -0.066), (34, 0.029), (35, 0.007), (36, -0.109), (37, 0.137), (38, -0.028), (39, -0.025), (40, -0.079), (41, 0.009), (42, -0.019), (43, 0.038), (44, 0.069), (45, 0.087), (46, 0.039), (47, 0.058), (48, 0.049), (49, -0.023)]
simIndex simValue paperId paperTitle
same-paper 1 0.93102801 163 acl-2011-Improved Modeling of Out-Of-Vocabulary Words Using Morphological Classes
Author: Thomas Mueller ; Hinrich Schuetze
Abstract: We present a class-based language model that clusters rare words of similar morphology together. The model improves the prediction of words after histories containing outof-vocabulary words. The morphological features used are obtained without the use of labeled data. The perplexity improvement compared to a state of the art Kneser-Ney model is 4% overall and 81% on unknown histories.
2 0.76734489 203 acl-2011-Learning Sub-Word Units for Open Vocabulary Speech Recognition
Author: Carolina Parada ; Mark Dredze ; Abhinav Sethy ; Ariya Rastrow
Abstract: Large vocabulary speech recognition systems fail to recognize words beyond their vocabulary, many of which are information rich terms, like named entities or foreign words. Hybrid word/sub-word systems solve this problem by adding sub-word units to large vocabulary word based systems; new words can then be represented by combinations of subword units. Previous work heuristically created the sub-word lexicon from phonetic representations of text using simple statistics to select common phone sequences. We propose a probabilistic model to learn the subword lexicon optimized for a given task. We consider the task of out of vocabulary (OOV) word detection, which relies on output from a hybrid model. A hybrid model with our learned sub-word lexicon reduces error by 6.3% and 7.6% (absolute) at a 5% false alarm rate on an English Broadcast News and MIT Lectures task respectively.
3 0.71615887 175 acl-2011-Integrating history-length interpolation and classes in language modeling
Author: Hinrich Schutze
Abstract: Building on earlier work that integrates different factors in language modeling, we view (i) backing off to a shorter history and (ii) class-based generalization as two complementary mechanisms of using a larger equivalence class for prediction when the default equivalence class is too small for reliable estimation. This view entails that the classes in a language model should be learned from rare events only and should be preferably applied to rare events. We construct such a model and show that both training on rare events and preferable application to rare events improve perplexity when compared to a simple direct interpolation of class-based with standard language models.
4 0.60731405 142 acl-2011-Generalized Interpolation in Decision Tree LM
Author: Denis Filimonov ; Mary Harper
Abstract: In the face of sparsity, statistical models are often interpolated with lower order (backoff) models, particularly in Language Modeling. In this paper, we argue that there is a relation between the higher order and the backoff model that must be satisfied in order for the interpolation to be effective. We show that in n-gram models, the relation is trivially held, but in models that allow arbitrary clustering of context (such as decision tree models), this relation is generally not satisfied. Based on this insight, we also propose a generalization of linear interpolation which significantly improves the performance of a decision tree language model.
5 0.60152858 38 acl-2011-An Empirical Investigation of Discounting in Cross-Domain Language Models
Author: Greg Durrett ; Dan Klein
Abstract: We investigate the empirical behavior of ngram discounts within and across domains. When a language model is trained and evaluated on two corpora from exactly the same domain, discounts are roughly constant, matching the assumptions of modified Kneser-Ney LMs. However, when training and test corpora diverge, the empirical discount grows essentially as a linear function of the n-gram count. We adapt a Kneser-Ney language model to incorporate such growing discounts, resulting in perplexity improvements over modified Kneser-Ney and Jelinek-Mercer baselines.
6 0.6006276 208 acl-2011-Lexical Normalisation of Short Text Messages: Makn Sens a #twitter
7 0.54701954 10 acl-2011-A Discriminative Model for Joint Morphological Disambiguation and Dependency Parsing
8 0.54150939 124 acl-2011-Exploiting Morphology in Turkish Named Entity Recognition System
9 0.52094722 24 acl-2011-A Scalable Probabilistic Classifier for Language Modeling
10 0.51681572 172 acl-2011-Insertion, Deletion, or Substitution? Normalizing Text Messages without Pre-categorization nor Supervision
11 0.47119626 319 acl-2011-Unsupervised Decomposition of a Document into Authorial Components
12 0.46640661 75 acl-2011-Combining Morpheme-based Machine Translation with Post-processing Morpheme Prediction
13 0.45933908 318 acl-2011-Unsupervised Bilingual Morpheme Segmentation and Alignment with Context-rich Hidden Semi-Markov Models
14 0.43978825 301 acl-2011-The impact of language models and loss functions on repair disfluency detection
15 0.43685868 15 acl-2011-A Hierarchical Pitman-Yor Process HMM for Unsupervised Part of Speech Induction
16 0.41737679 193 acl-2011-Language-independent compound splitting with morphological operations
17 0.41487348 238 acl-2011-P11-2093 k2opt.pdf
18 0.41015476 140 acl-2011-Fully Unsupervised Word Segmentation with BVE and MDL
19 0.39493802 329 acl-2011-Using Deep Morphology to Improve Automatic Error Detection in Arabic Handwriting Recognition
20 0.38549241 310 acl-2011-Translating from Morphologically Complex Languages: A Paraphrase-Based Approach
topicId topicWeight
[(5, 0.027), (17, 0.056), (26, 0.025), (31, 0.033), (37, 0.062), (39, 0.049), (41, 0.046), (45, 0.237), (55, 0.09), (59, 0.025), (72, 0.04), (91, 0.03), (96, 0.178), (97, 0.027)]
simIndex simValue paperId paperTitle
Author: Roger Levy
Abstract: A system making optimal use of available information in incremental language comprehension might be expected to use linguistic knowledge together with current input to revise beliefs about previous input. Under some circumstances, such an error-correction capability might induce comprehenders to adopt grammatical analyses that are inconsistent with the true input. Here we present a formal model of how such input-unfaithful garden paths may be adopted and the difficulty incurred by their subsequent disconfirmation, combining a rational noisy-channel model of syntactic comprehension under uncertain input with the surprisal theory of incremental processing difficulty. We also present a behavioral experiment confirming the key empirical predictions of the theory.
same-paper 2 0.78495997 163 acl-2011-Improved Modeling of Out-Of-Vocabulary Words Using Morphological Classes
Author: Thomas Mueller ; Hinrich Schuetze
Abstract: We present a class-based language model that clusters rare words of similar morphology together. The model improves the prediction of words after histories containing outof-vocabulary words. The morphological features used are obtained without the use of labeled data. The perplexity improvement compared to a state of the art Kneser-Ney model is 4% overall and 81% on unknown histories.
3 0.7695241 98 acl-2011-Discovery of Topically Coherent Sentences for Extractive Summarization
Author: Asli Celikyilmaz ; Dilek Hakkani-Tur
Abstract: Extractive methods for multi-document summarization are mainly governed by information overlap, coherence, and content constraints. We present an unsupervised probabilistic approach to model the hidden abstract concepts across documents as well as the correlation between these concepts, to generate topically coherent and non-redundant summaries. Based on human evaluations our models generate summaries with higher linguistic quality in terms of coherence, readability, and redundancy compared to benchmark systems. Although our system is unsupervised and optimized for topical coherence, we achieve a 44.1 ROUGE on the DUC-07 test set, roughly in the range of state-of-the-art supervised models.
4 0.74973619 90 acl-2011-Crowdsourcing Translation: Professional Quality from Non-Professionals
Author: Omar F. Zaidan ; Chris Callison-Burch
Abstract: Naively collecting translations by crowdsourcing the task to non-professional translators yields disfluent, low-quality results if no quality control is exercised. We demonstrate a variety of mechanisms that increase the translation quality to near professional levels. Specifically, we solicit redundant translations and edits to them, and automatically select the best output among them. We propose a set of features that model both the translations and the translators, such as country of residence, LM perplexity of the translation, edit rate from the other translations, and (optionally) calibration against professional translators. Using these features to score the collected translations, we are able to discriminate between acceptable and unacceptable translations. We recreate the NIST 2009 Urdu-toEnglish evaluation set with Mechanical Turk, and quantitatively show that our models are able to select translations within the range of quality that we expect from professional trans- lators. The total cost is more than an order of magnitude lower than professional translation.
5 0.67804742 144 acl-2011-Global Learning of Typed Entailment Rules
Author: Jonathan Berant ; Ido Dagan ; Jacob Goldberger
Abstract: Extensive knowledge bases ofentailment rules between predicates are crucial for applied semantic inference. In this paper we propose an algorithm that utilizes transitivity constraints to learn a globally-optimal set of entailment rules for typed predicates. We model the task as a graph learning problem and suggest methods that scale the algorithm to larger graphs. We apply the algorithm over a large data set of extracted predicate instances, from which a resource of typed entailment rules has been recently released (Schoenmackers et al., 2010). Our results show that using global transitivity information substantially improves performance over this resource and several baselines, and that our scaling methods allow us to increase the scope of global learning of entailment-rule graphs.
6 0.67613089 175 acl-2011-Integrating history-length interpolation and classes in language modeling
8 0.67108214 237 acl-2011-Ordering Prenominal Modifiers with a Reranking Approach
9 0.66983312 245 acl-2011-Phrase-Based Translation Model for Question Retrieval in Community Question Answer Archives
10 0.66923362 124 acl-2011-Exploiting Morphology in Turkish Named Entity Recognition System
11 0.66823268 318 acl-2011-Unsupervised Bilingual Morpheme Segmentation and Alignment with Context-rich Hidden Semi-Markov Models
12 0.66711259 275 acl-2011-Semi-Supervised Modeling for Prenominal Modifier Ordering
13 0.66683823 15 acl-2011-A Hierarchical Pitman-Yor Process HMM for Unsupervised Part of Speech Induction
14 0.66676742 171 acl-2011-Incremental Syntactic Language Models for Phrase-based Translation
15 0.66654396 14 acl-2011-A Hierarchical Model of Web Summaries
16 0.66542053 38 acl-2011-An Empirical Investigation of Discounting in Cross-Domain Language Models
17 0.66488683 280 acl-2011-Sentence Ordering Driven by Local and Global Coherence for Summary Generation
18 0.66171402 32 acl-2011-Algorithm Selection and Model Adaptation for ESL Correction Tasks
19 0.66118032 137 acl-2011-Fine-Grained Class Label Markup of Search Queries
20 0.65984225 44 acl-2011-An exponential translation model for target language morphology