emnlp emnlp2010 emnlp2010-99 knowledge-graph by maker-knowledge-mining
Source: pdf
Author: Libin Shen ; Bing Zhang ; Spyros Matsoukas ; Jinxi Xu ; Ralph Weischedel
Abstract: In modern machine translation practice, a statistical phrasal or hierarchical translation system usually relies on a huge set of translation rules extracted from bi-lingual training data. This approach not only results in space and efficiency issues, but also suffers from the sparse data problem. In this paper, we propose to use factorized grammars, an idea widely accepted in the field of linguistic grammar construction, to generalize translation rules, so as to solve these two problems. We designed a method to take advantage of the XTAG English Grammar to facilitate the extraction of factorized rules. We experimented on various setups of low-resource language translation, and showed consistent significant improvement in BLEU over state-ofthe-art string-to-dependency baseline systems with 200K words of bi-lingual training data.
Reference: text
sentIndex sentText sentNum sentScore
1 com s , , , Abstract In modern machine translation practice, a statistical phrasal or hierarchical translation system usually relies on a huge set of translation rules extracted from bi-lingual training data. [sent-2, score-0.842]
2 In this paper, we propose to use factorized grammars, an idea widely accepted in the field of linguistic grammar construction, to generalize translation rules, so as to solve these two problems. [sent-4, score-0.827]
3 We designed a method to take advantage of the XTAG English Grammar to facilitate the extraction of factorized rules. [sent-5, score-0.486]
4 , 2006) machine translation system usually relies on a very large set of translation rules extracted from bi-lingual training data with heuristic methods on word alignment results. [sent-9, score-0.54]
5 A common practice to circumvent this problem is to filter the rules based on development sets in the step of rule extraction or before the decoding phrase, instead of building a real distributed system. [sent-12, score-0.267]
6 However, this strategy only works for research systems, 616 for which the segments for translation are always fixed. [sent-13, score-0.237]
7 Linguists decompose lexicalized linguistic structures into two parts, (unlexicalized) templates and lexical items. [sent-17, score-0.483]
8 Each family is associated with a set of lexical items which can be used to lexicalize all the templates in this family. [sent-19, score-0.694]
9 For example, the XTAG English Grammar (XTAG-Group, 2001), a hand-crafted grammar based on the Tree Adjoining Grammar (TAG) (Joshi and Schabes, 1997) formalism, is a grammar of this kind, which employs factorization with LTAG e-tree templates and lexical items. [sent-20, score-0.645]
10 Factorized grammars not only relieve the burden on space and search, but also alleviate the sparse data problem, especially for low-resource language translation with few training data. [sent-21, score-0.425]
11 New rules can be generated from template families and lexical items either offline or on the fly, explicitly or implicitly. [sent-23, score-0.856]
12 In this work, we will go further to investigate factorization of rule structures by exploiting the rich XTAG English Grammar. [sent-25, score-0.202]
13 We evaluate the effect of using factorized translation grammars on various setups of low-resource language translation, since low-resource MT suffers greatly on poor generalization capability of transProceMedITin,g Ms oasfs thaceh 2u0se1t0ts C,o UnSfAer,e n9c-e1 on O Ectmobpeir ic 2a0l1 M0. [sent-26, score-0.915]
14 With the help of high-level linguistic knowledge for generalization, factorized grammars provide consistent significant improvement in BLEU (Papineni et al. [sent-29, score-0.64]
15 This work also closes the gap between compact hand-crafted translation rules and large-scale unorganized automatic rules. [sent-31, score-0.325]
16 This may lead to a more effective and efficient statistical translation model that could better leverage generic linguistic knowledge in MT. [sent-32, score-0.329]
17 Then, we will introduce factorized translation grammars in Section 3. [sent-34, score-0.772]
18 We will illustrate the use of the XTAG English Grammar to facilitate the extraction of factorized rules in Section 4. [sent-35, score-0.596]
19 First, it uses syntactic tree structures on the target side, which makes it easy to exploit linguistic information. [sent-41, score-0.183]
20 Third, a string-to-dependency system provides state-of-theart performance on translation accuracy, so that improvement over such a system will be more convincing. [sent-43, score-0.26]
21 In the baseline string-to-dependency model, each translation rule is composed of two parts, source and target. [sent-48, score-0.401]
22 The source sides is a string rewriting rule, and the target side is a tree rewriting rule. [sent-49, score-0.322]
23 The source side of a translation rule is used to detect when this rule can be applied. [sent-53, score-0.601]
24 The target side of the rule provides a hypothesis tree structure for the matched span. [sent-54, score-0.294]
25 Mono-lingual parsing can be viewed as a special case of this generic algorithm, for which the source string is a projection of the target tree structure. [sent-55, score-0.199]
26 Figure 1 shows three examples of string-todependency translation rules. [sent-56, score-0.215]
27 In Figure 1, the top boxes mean the source side, and the bottom boxes mean the target side. [sent-60, score-0.208]
28 There is no rule having both HATE and FUN Q on its source side. [sent-70, score-0.186]
29 Intuitively, we would like to have translation rule that tell us how to translate X1 HATE X2 FUN Q as in Figure 2. [sent-74, score-0.377]
30 Furthermore, if we know like and hate are in the same syntactic/semantic class in the source or target language, we will be very confident on the validity of this hypothesis rule. [sent-77, score-0.346]
31 Now, we propose a factorized grammar to solve this generalization problem. [sent-78, score-0.612]
32 In addition, translation rules represented with the new formalism will be more compact. [sent-79, score-0.361]
33 1 Factorized Rules We decompose a translation rule into two parts, a pair of lexical items and an unlexicalized template. [sent-81, score-0.648]
34 It is similar to the solution in the XTAG English Grammar (XTAG-Group, 2001), while here we Figure 1: Three examples of string-to-dependency translation rules. [sent-82, score-0.215]
35 Thus, the original rule becomes an unlexicalized rule template. [sent-88, score-0.301]
36 As for the three example rules in Figure 1, we will 618 extract lexical items (LIKE, like), (HATE, hate) and (LIKE, like) respectively. [sent-89, score-0.332]
37 We obtain the same lexical items from the first and the third rules. [sent-90, score-0.222]
38 Here, V represents a verb on the source side, VB stands for a verb in the base form, and VBZ means a verb in the third person singular present form as in the Penn Treebank representation (Marcus et al. [sent-92, score-0.211]
39 In the XTAG English Grammar, tree templates for transitive verbs are grouped into a family. [sent-94, score-0.444]
40 Here, we assume that the rule templates representing structural variations of the same word class can also be organized into a template family. [sent-96, score-0.574]
41 For example, as shown in Figure 4, templates and lexical items are associated with families. [sent-97, score-0.559]
42 It should be noted that a template or a lexical item can be associated with more than one family. [sent-98, score-0.258]
43 As for the missing Figure 4: Templates and lexical items are associated with families. [sent-100, score-0.273]
44 rule in Figure 2, we can now generate it by replacing the POS tags in the second template of Figure 4 with lexical items (HATE, hate) with their correct inflections. [sent-101, score-0.486]
45 Both the template and the lexical items here are associated with the family Transitive 3. [sent-102, score-0.515]
46 We decompose a rule R into to two parts, its template PR and its lexical items LR. [sent-106, score-0.529]
47 The second generative model will also be useful for unsupervised learning of families and related probabilities. [sent-109, score-0.333]
48 In this paper, we approximate families by using target (English) side linguistic knowledge as what we will explain in Section 4, so this changes the definition of the task. [sent-110, score-0.493]
49 We will also be given an association table B(L, F) for lexical items L and families F, such 619 that B(L, F) = true if and only L is associated with F, but we do not know the distributions. [sent-112, score-0.582]
50 Let S be the source side of a rule or a rule template, T the target side of a rule of a rule template. [sent-113, score-0.774]
51 Let P and L be the template and lexical items of R respectively. [sent-116, score-0.353]
52 3 Discussion The factorized models discussed in the previous section can greatly alleviate the sparse data problem, especially for low-resource translation tasks. [sent-124, score-0.754]
53 Therefore, to use unsupervised learning with a model like (1) somehow reduces a hard translation problem to another one of the same difficulty, when the training data is small. [sent-126, score-0.215]
54 For example, if the target language has rich resources, although the source language is a low-density one, we can exploit the linguistic knowledge on the target side, and carry it over to bi-lingual structures of the translation model. [sent-128, score-0.467]
55 The setup of X-to-English translation tasks is just like this. [sent-129, score-0.215]
56 We leave unsupervised learning of factorized translation grammars for future research. [sent-131, score-0.772]
57 4 Using A Mono-Lingual Grammar In this section, we will focus on X-to-English translation, and explain how to use English resources to build a factorized translation grammar. [sent-132, score-0.705]
58 As shown in Figure 4, intuitively, the families are intersection of the word families of the two languages involved, which means that they are refinement of the English word families. [sent-134, score-0.666]
59 For example, a sub-set of the English transitive families may be translated in the same way, so they share the same set of templates. [sent-135, score-0.428]
60 This is why we named the two families Transitive 3 and Intransitive 2 in Figure 4. [sent-136, score-0.333]
61 Therefore, we approximate bi-lingual families with English families first. [sent-137, score-0.666]
62 About 30,000 lexical items are associ- ated with these families and individual templates 1. [sent-149, score-0.865]
63 In addition, it also has the richest English morphological lexicon with 3 17,000 inflected items derived from 90,000 stems. [sent-150, score-0.184]
64 In our applications, we select all the verb families plus one each for nouns, adjectives and adverbs. [sent-152, score-0.39]
65 We use the families of the English word as the families of bi-lingual lexical items. [sent-153, score-0.73]
66 Therefore, we have a list of about 20 families and an association table as described in Section 3. [sent-154, score-0.333]
67 Of course, one can use other linguistic resources if similar family information is provided, e. [sent-156, score-0.204]
68 5 Implementation Nowadays, machine translation systems become more and more complicated. [sent-160, score-0.215]
69 A common practice is to reduce a new translation model to an old one, so that we can use an existing system, and see the effect of the new model quickly. [sent-162, score-0.215]
70 Instead of building a new decoder that uses factorized grammars, we reduce factorized rules to baseline string-todependency rules by performing combination of templates and lexical items in an offline mode. [sent-165, score-1.779]
71 In the rule extraction phase, we first extract all the string-to-dependency rules with the baseline system. [sent-168, score-0.243]
72 For each extracted rule, we try to split it into various “template–lexical item” pairs by choosing different aligned words for delexicalization, which turns rules in Figure 1 into lexical items and templates in Figure 3. [sent-174, score-0.642]
73 Events of templates and lexical items are counted according to the family of the target English word. [sent-175, score-0.722]
74 Then, we apply them to all the lexical items in this families, and save the generated rules. [sent-179, score-0.222]
75 Obviously, using only the 200 most frequent templates for each family is just a rough approximation. [sent-182, score-0.475]
76 An exact implementation of a new decoder for factorized grammars can make better use of all the templates. [sent-183, score-0.606]
77 Since we implement template application in an offline mode, we can use exactly the same decoding and optimization algorithms as the baseline. [sent-187, score-0.215]
78 The decoder is a generic chart parsing algorithm that generates target dependency trees from source string input. [sent-188, score-0.232]
79 6 Experiments on Low-Resource Setups We tested the performance of using factorized grammars on low-resource MT setups. [sent-190, score-0.557]
80 This is one of the cases that a factorized grammar would help. [sent-192, score-0.574]
81 Chinese-to-English translation is from a morphology poor language to a morphology rich language, while Arabic-to-English translation is in the opposite direction. [sent-212, score-0.43]
82 It will be interesting to see if factorized grammars help on both cases. [sent-213, score-0.58]
83 The system using factorized grammars shows BLEU improvement in all conditions. [sent-222, score-0.602]
84 The reason why the improvement diminishes on the full data set could be that the rough approximation with 200 most frequent templates cannot fully take advantage of this paradigm, which will be discussed in the next section. [sent-239, score-0.385]
85 data size in logscale Figure 6: BLEU Improvement with 95% confidence range by using factorized grammars on TEST-1 . [sent-242, score-0.605]
86 3 Example Templates Figure 8 lists seven Arabic-to-English templates randomly selected from the transitive verb family. [sent-244, score-0.437]
87 However, we notice that most of the templates in the 200 lists are rather simple. [sent-247, score-0.31]
88 623 data size in logscale Figure 7: BLEU Improvement with 95% confidence range by using factorized grammars on TEST-2. [sent-250, score-0.605]
89 Generic rules like this will be very close to handcraft translate rules that people have accumulated for rule-based MT systems. [sent-251, score-0.249]
90 7 Conclusions and Future Work In this paper, we proposed a novel statistical machine translation model using a factorized structurebased translation grammar. [sent-252, score-0.913]
91 TMPL_1 TMPL_31 VX1 V VBX1VBD TMPL_121 X1 V TMPL_61 TMPL_91 TMPL_151 TMPL_181 X1 V X2 X1VBZX2 Figure 8: Randomly selected Arabic-to-English templates from the transitive verb family. [sent-254, score-0.437]
92 We took low-resource language translation, especially X-to-English translation tasks, for case study. [sent-255, score-0.215]
93 We designed a method to exploit family information in the XTAG English Grammar to facilitate the extraction of factorized rules. [sent-256, score-0.644]
94 We tested the new model on low-resource translation, and the use of factorized models showed significant improvement in BLEU on systems with 200K words of bi-lingual training data of various language pairs and genres. [sent-257, score-0.504]
95 The factorized translation grammar proposed here shows an interesting way of using richer syntactic resources, with high potential for future research. [sent-258, score-0.812]
96 In future, we will explore various learning methods for better estimation of families, templates and lexical items. [sent-259, score-0.374]
97 624 We will also try to further exploit the factorized representation with discriminative learning. [sent-261, score-0.482]
98 Features defined on templates and families will have good generalization capability. [sent-262, score-0.681]
99 SPMT: Statistical machine translation with syntactified target language phrases. [sent-321, score-0.27]
100 A new string-to-dependency machine translation algorithm with a target dependency language model. [sent-343, score-0.27]
wordName wordTfidf (topN-words)
[('factorized', 0.459), ('xtag', 0.333), ('families', 0.333), ('templates', 0.31), ('hate', 0.238), ('translation', 0.215), ('items', 0.158), ('bleu', 0.142), ('family', 0.135), ('rule', 0.133), ('template', 0.131), ('pr', 0.12), ('grammar', 0.115), ('rules', 0.11), ('grammars', 0.098), ('transitive', 0.095), ('meteor', 0.085), ('setups', 0.073), ('fun', 0.073), ('prb', 0.071), ('bl', 0.068), ('side', 0.067), ('lexical', 0.064), ('stands', 0.062), ('mt', 0.06), ('offline', 0.06), ('english', 0.055), ('target', 0.055), ('source', 0.053), ('generic', 0.052), ('jinxi', 0.051), ('decoder', 0.049), ('indirection', 0.048), ('logscale', 0.048), ('prt', 0.048), ('deneefe', 0.048), ('libin', 0.048), ('shen', 0.047), ('sparse', 0.047), ('improvement', 0.045), ('adjoining', 0.045), ('sides', 0.044), ('xu', 0.044), ('decompose', 0.043), ('kipper', 0.041), ('joshi', 0.041), ('factorization', 0.041), ('phrasal', 0.04), ('rs', 0.039), ('tree', 0.039), ('linguistic', 0.038), ('generalization', 0.038), ('rt', 0.037), ('ralph', 0.037), ('item', 0.036), ('formalism', 0.036), ('koehn', 0.035), ('unlexicalized', 0.035), ('boxes', 0.034), ('banerjee', 0.034), ('bbn', 0.034), ('matsoukas', 0.034), ('spyros', 0.034), ('aravind', 0.034), ('alleviate', 0.033), ('suffers', 0.032), ('verb', 0.032), ('linguists', 0.032), ('burden', 0.032), ('rewriting', 0.032), ('bottom', 0.032), ('resources', 0.031), ('rough', 0.03), ('translate', 0.029), ('tune', 0.029), ('ps', 0.028), ('lr', 0.028), ('sake', 0.028), ('structures', 0.028), ('associated', 0.027), ('met', 0.027), ('facilitate', 0.027), ('figure', 0.027), ('fellbaum', 0.026), ('gale', 0.026), ('morphological', 0.026), ('plus', 0.025), ('factored', 0.025), ('smoothed', 0.025), ('decoding', 0.024), ('missing', 0.024), ('statistical', 0.024), ('hierarchical', 0.023), ('exploit', 0.023), ('chart', 0.023), ('phase', 0.023), ('carreras', 0.023), ('interesting', 0.023), ('och', 0.023), ('segments', 0.022)]
simIndex simValue paperId paperTitle
same-paper 1 0.9999997 99 emnlp-2010-Statistical Machine Translation with a Factorized Grammar
Author: Libin Shen ; Bing Zhang ; Spyros Matsoukas ; Jinxi Xu ; Ralph Weischedel
Abstract: In modern machine translation practice, a statistical phrasal or hierarchical translation system usually relies on a huge set of translation rules extracted from bi-lingual training data. This approach not only results in space and efficiency issues, but also suffers from the sparse data problem. In this paper, we propose to use factorized grammars, an idea widely accepted in the field of linguistic grammar construction, to generalize translation rules, so as to solve these two problems. We designed a method to take advantage of the XTAG English Grammar to facilitate the extraction of factorized rules. We experimented on various setups of low-resource language translation, and showed consistent significant improvement in BLEU over state-ofthe-art string-to-dependency baseline systems with 200K words of bi-lingual training data.
2 0.20891035 57 emnlp-2010-Hierarchical Phrase-Based Translation Grammars Extracted from Alignment Posterior Probabilities
Author: Adria de Gispert ; Juan Pino ; William Byrne
Abstract: We report on investigations into hierarchical phrase-based translation grammars based on rules extracted from posterior distributions over alignments of the parallel text. Rather than restrict rule extraction to a single alignment, such as Viterbi, we instead extract rules based on posterior distributions provided by the HMM word-to-word alignmentmodel. We define translation grammars progressively by adding classes of rules to a basic phrase-based system. We assess these grammars in terms of their expressive power, measured by their ability to align the parallel text from which their rules are extracted, and the quality of the translations they yield. In Chinese-to-English translation, we find that rule extraction from posteriors gives translation improvements. We also find that grammars with rules with only one nonterminal, when extracted from posteri- ors, can outperform more complex grammars extracted from Viterbi alignments. Finally, we show that the best way to exploit source-totarget and target-to-source alignment models is to build two separate systems and combine their output translation lattices.
3 0.15016952 33 emnlp-2010-Cross Language Text Classification by Model Translation and Semi-Supervised Learning
Author: Lei Shi ; Rada Mihalcea ; Mingjun Tian
Abstract: In this paper, we introduce a method that automatically builds text classifiers in a new language by training on already labeled data in another language. Our method transfers the classification knowledge across languages by translating the model features and by using an Expectation Maximization (EM) algorithm that naturally takes into account the ambiguity associated with the translation of a word. We further exploit the readily available unlabeled data in the target language via semisupervised learning, and adapt the translated model to better fit the data distribution of the target language.
4 0.1377205 86 emnlp-2010-Non-Isomorphic Forest Pair Translation
Author: Hui Zhang ; Min Zhang ; Haizhou Li ; Eng Siong Chng
Abstract: This paper studies two issues, non-isomorphic structure translation and target syntactic structure usage, for statistical machine translation in the context of forest-based tree to tree sequence translation. For the first issue, we propose a novel non-isomorphic translation framework to capture more non-isomorphic structure mappings than traditional tree-based and tree-sequence-based translation methods. For the second issue, we propose a parallel space searching method to generate hypothesis using tree-to-string model and evaluate its syntactic goodness using tree-to-tree/tree sequence model. This not only reduces the search complexity by merging spurious-ambiguity translation paths and solves the data sparseness issue in training, but also serves as a syntax-based target language model for better grammatical generation. Experiment results on the benchmark data show our proposed two solutions are very effective, achieving significant performance improvement over baselines when applying to different translation models.
5 0.13513678 63 emnlp-2010-Improving Translation via Targeted Paraphrasing
Author: Philip Resnik ; Olivia Buzek ; Chang Hu ; Yakov Kronrod ; Alex Quinn ; Benjamin B. Bederson
Abstract: Targeted paraphrasing is a new approach to the problem of obtaining cost-effective, reasonable quality translation that makes use of simple and inexpensive human computations by monolingual speakers in combination with machine translation. The key insight behind the process is that it is possible to spot likely translation errors with only monolingual knowledge of the target language, and it is possible to generate alternative ways to say the same thing (i.e. paraphrases) with only monolingual knowledge of the source language. Evaluations demonstrate that this approach can yield substantial improvements in translation quality.
6 0.12548457 29 emnlp-2010-Combining Unsupervised and Supervised Alignments for MT: An Empirical Study
7 0.12399708 69 emnlp-2010-Joint Training and Decoding Using Virtual Nodes for Cascaded Segmentation and Tagging Tasks
8 0.12254933 78 emnlp-2010-Minimum Error Rate Training by Sampling the Translation Lattice
10 0.11667896 47 emnlp-2010-Example-Based Paraphrasing for Improved Phrase-Based Statistical Machine Translation
11 0.1124296 5 emnlp-2010-A Hybrid Morpheme-Word Representation for Machine Translation of Morphologically Rich Languages
12 0.11063717 18 emnlp-2010-Assessing Phrase-Based Translation Models with Oracle Decoding
13 0.10531236 50 emnlp-2010-Facilitating Translation Using Source Language Paraphrase Lattices
14 0.095388979 76 emnlp-2010-Maximum Entropy Based Phrase Reordering for Hierarchical Phrase-Based Translation
15 0.093803987 96 emnlp-2010-Self-Training with Products of Latent Variable Grammars
16 0.092115439 22 emnlp-2010-Automatic Evaluation of Translation Quality for Distant Language Pairs
17 0.089676835 42 emnlp-2010-Efficient Incremental Decoding for Tree-to-String Translation
18 0.079859197 36 emnlp-2010-Discriminative Word Alignment with a Function Word Reordering Model
19 0.076552667 65 emnlp-2010-Inducing Probabilistic CCG Grammars from Logical Form with Higher-Order Unification
20 0.073573545 39 emnlp-2010-EMNLP 044
topicId topicWeight
[(0, 0.284), (1, -0.203), (2, 0.089), (3, -0.058), (4, -0.011), (5, -0.016), (6, 0.002), (7, -0.05), (8, -0.129), (9, -0.031), (10, 0.077), (11, -0.054), (12, -0.002), (13, -0.079), (14, -0.098), (15, -0.005), (16, -0.027), (17, 0.019), (18, -0.015), (19, -0.053), (20, 0.067), (21, -0.237), (22, 0.026), (23, -0.041), (24, 0.143), (25, -0.058), (26, 0.08), (27, -0.084), (28, -0.113), (29, 0.07), (30, 0.01), (31, 0.115), (32, -0.038), (33, -0.035), (34, 0.094), (35, -0.155), (36, 0.015), (37, 0.009), (38, -0.108), (39, 0.178), (40, -0.051), (41, 0.094), (42, -0.064), (43, -0.16), (44, -0.069), (45, -0.051), (46, -0.067), (47, -0.017), (48, 0.058), (49, 0.037)]
simIndex simValue paperId paperTitle
same-paper 1 0.9306345 99 emnlp-2010-Statistical Machine Translation with a Factorized Grammar
Author: Libin Shen ; Bing Zhang ; Spyros Matsoukas ; Jinxi Xu ; Ralph Weischedel
Abstract: In modern machine translation practice, a statistical phrasal or hierarchical translation system usually relies on a huge set of translation rules extracted from bi-lingual training data. This approach not only results in space and efficiency issues, but also suffers from the sparse data problem. In this paper, we propose to use factorized grammars, an idea widely accepted in the field of linguistic grammar construction, to generalize translation rules, so as to solve these two problems. We designed a method to take advantage of the XTAG English Grammar to facilitate the extraction of factorized rules. We experimented on various setups of low-resource language translation, and showed consistent significant improvement in BLEU over state-ofthe-art string-to-dependency baseline systems with 200K words of bi-lingual training data.
2 0.63171738 57 emnlp-2010-Hierarchical Phrase-Based Translation Grammars Extracted from Alignment Posterior Probabilities
Author: Adria de Gispert ; Juan Pino ; William Byrne
Abstract: We report on investigations into hierarchical phrase-based translation grammars based on rules extracted from posterior distributions over alignments of the parallel text. Rather than restrict rule extraction to a single alignment, such as Viterbi, we instead extract rules based on posterior distributions provided by the HMM word-to-word alignmentmodel. We define translation grammars progressively by adding classes of rules to a basic phrase-based system. We assess these grammars in terms of their expressive power, measured by their ability to align the parallel text from which their rules are extracted, and the quality of the translations they yield. In Chinese-to-English translation, we find that rule extraction from posteriors gives translation improvements. We also find that grammars with rules with only one nonterminal, when extracted from posteri- ors, can outperform more complex grammars extracted from Viterbi alignments. Finally, we show that the best way to exploit source-totarget and target-to-source alignment models is to build two separate systems and combine their output translation lattices.
3 0.49663493 29 emnlp-2010-Combining Unsupervised and Supervised Alignments for MT: An Empirical Study
Author: Jinxi Xu ; Antti-Veikko Rosti
Abstract: Word alignment plays a central role in statistical MT (SMT) since almost all SMT systems extract translation rules from word aligned parallel training data. While most SMT systems use unsupervised algorithms (e.g. GIZA++) for training word alignment, supervised methods, which exploit a small amount of human-aligned data, have become increasingly popular recently. This work empirically studies the performance of these two classes of alignment algorithms and explores strategies to combine them to improve overall system performance. We used two unsupervised aligners, GIZA++ and HMM, and one supervised aligner, ITG, in this study. To avoid language and genre specific conclusions, we ran experiments on test sets consisting of two language pairs (Chinese-to-English and Arabicto-English) and two genres (newswire and weblog). Results show that the two classes of algorithms achieve the same level of MT perfor- mance. Modest improvements were achieved by taking the union of the translation grammars extracted from different alignments. Significant improvements (around 1.0 in BLEU) were achieved by combining outputs of different systems trained with different alignments. The improvements are consistent across languages and genres.
4 0.48170704 86 emnlp-2010-Non-Isomorphic Forest Pair Translation
Author: Hui Zhang ; Min Zhang ; Haizhou Li ; Eng Siong Chng
Abstract: This paper studies two issues, non-isomorphic structure translation and target syntactic structure usage, for statistical machine translation in the context of forest-based tree to tree sequence translation. For the first issue, we propose a novel non-isomorphic translation framework to capture more non-isomorphic structure mappings than traditional tree-based and tree-sequence-based translation methods. For the second issue, we propose a parallel space searching method to generate hypothesis using tree-to-string model and evaluate its syntactic goodness using tree-to-tree/tree sequence model. This not only reduces the search complexity by merging spurious-ambiguity translation paths and solves the data sparseness issue in training, but also serves as a syntax-based target language model for better grammatical generation. Experiment results on the benchmark data show our proposed two solutions are very effective, achieving significant performance improvement over baselines when applying to different translation models.
5 0.48124263 33 emnlp-2010-Cross Language Text Classification by Model Translation and Semi-Supervised Learning
Author: Lei Shi ; Rada Mihalcea ; Mingjun Tian
Abstract: In this paper, we introduce a method that automatically builds text classifiers in a new language by training on already labeled data in another language. Our method transfers the classification knowledge across languages by translating the model features and by using an Expectation Maximization (EM) algorithm that naturally takes into account the ambiguity associated with the translation of a word. We further exploit the readily available unlabeled data in the target language via semisupervised learning, and adapt the translated model to better fit the data distribution of the target language.
6 0.45525247 5 emnlp-2010-A Hybrid Morpheme-Word Representation for Machine Translation of Morphologically Rich Languages
7 0.45077902 96 emnlp-2010-Self-Training with Products of Latent Variable Grammars
8 0.4496555 13 emnlp-2010-A Simple Domain-Independent Probabilistic Approach to Generation
9 0.44270477 63 emnlp-2010-Improving Translation via Targeted Paraphrasing
10 0.43945238 78 emnlp-2010-Minimum Error Rate Training by Sampling the Translation Lattice
11 0.42115882 98 emnlp-2010-Soft Syntactic Constraints for Hierarchical Phrase-Based Translation Using Latent Syntactic Distributions
12 0.42003679 22 emnlp-2010-Automatic Evaluation of Translation Quality for Distant Language Pairs
13 0.39885712 18 emnlp-2010-Assessing Phrase-Based Translation Models with Oracle Decoding
14 0.38468805 76 emnlp-2010-Maximum Entropy Based Phrase Reordering for Hierarchical Phrase-Based Translation
15 0.34146389 113 emnlp-2010-Unsupervised Induction of Tree Substitution Grammars for Dependency Parsing
16 0.34063485 69 emnlp-2010-Joint Training and Decoding Using Virtual Nodes for Cascaded Segmentation and Tagging Tasks
17 0.3326216 42 emnlp-2010-Efficient Incremental Decoding for Tree-to-String Translation
18 0.32316583 35 emnlp-2010-Discriminative Sample Selection for Statistical Machine Translation
19 0.32296368 72 emnlp-2010-Learning First-Order Horn Clauses from Web Text
20 0.32096261 50 emnlp-2010-Facilitating Translation Using Source Language Paraphrase Lattices
topicId topicWeight
[(10, 0.014), (12, 0.028), (29, 0.612), (32, 0.015), (52, 0.035), (56, 0.033), (62, 0.015), (66, 0.094), (72, 0.027), (76, 0.023), (87, 0.011)]
simIndex simValue paperId paperTitle
1 0.97517687 52 emnlp-2010-Further Meta-Evaluation of Broad-Coverage Surface Realization
Author: Dominic Espinosa ; Rajakrishnan Rajkumar ; Michael White ; Shoshana Berleant
Abstract: We present the first evaluation of the utility of automatic evaluation metrics on surface realizations of Penn Treebank data. Using outputs of the OpenCCG and XLE realizers, along with ranked WordNet synonym substitutions, we collected a corpus of generated surface realizations. These outputs were then rated and post-edited by human annotators. We evaluated the realizations using seven automatic metrics, and analyzed correlations obtained between the human judgments and the automatic scores. In contrast to previous NLG meta-evaluations, we find that several of the metrics correlate moderately well with human judgments of both adequacy and fluency, with the TER family performing best overall. We also find that all of the metrics correctly predict more than half of the significant systemlevel differences, though none are correct in all cases. We conclude with a discussion ofthe implications for the utility of such metrics in evaluating generation in the presence of variation. A further result of our research is a corpus of post-edited realizations, which will be made available to the research community. 1 Introduction and Background In building surface-realization systems for natural language generation, there is a need for reliable automated metrics to evaluate the output. Unlike in parsing, where there is usually a single goldstandard parse for a sentence, in surface realization there are usually many grammatically-acceptable ways to express the same concept. This parallels the task of evaluating machine-translation (MT) systems: for a given segment in the source language, 564 there are usually several acceptable translations into the target language. As human evaluation of translation quality is time-consuming and expensive, a number of automated metrics have been developed to evaluate the quality of MT outputs. In this study, we investigate whether the metrics developed for MT evaluation tasks can be used to reliably evaluate the outputs of surface realizers, and which of these metrics are best suited to this task. A number of surface realizers have been developed using the Penn Treebank (PTB), and BLEU scores are often reported in the evaluations of these systems. But how useful is BLEU in this context? The original BLEU study (Papineni et al., 2001) scored MT outputs, which are of generally lower quality than grammar-based surface realizations. Furthermore, even for MT systems, the usefulness of BLEU has been called into question (Callison-Burch et al., 2006). BLEU is designed to work with multiple reference sentences, but in treebank realization, there is only a single reference sentence available for comparison. A few other studies have investigated the use of such metrics in evaluating the output of NLG systems, notably (Reiter and Belz, 2009) and (Stent et al., 2005). The former examined the performance of BLEU and ROUGE with computer-generated weather reports, finding a moderate correlation with human fluency judgments. The latter study applied several MT metrics to paraphrase data from Barzilay and Lee’s corpus-based system (Barzilay and Lee, 2003), and found moderate correlations with human adequacy judgments, but little correlation with fluency judgments. Cahill (2009) examined the performance of six MT metrics (including BLEU) in evaluating the output of a LFG-based surface realizer for ProceMedITin,g Ms oasfs thaceh 2u0se1t0ts C,o UnSfAer,e n9c-e1 on O Ectmobpeir ic 2a0l1 M0.e ?tc ho2d0s10 in A Nsastoucira tlio Lnan fogru Cagoem Ppruotcaetisosninagl, L pinag eusis 5t6ic4s–574, German, also finding only weak correlations with the human judgments. To study the usefulness of evaluation metrics such as BLEU on the output of grammar-based surface realizers used with the PTB, we assembled a corpus of surface realizations from three different realizers operating on Section 00 of the PTB. Two human judges evaluated the adequacy and fluency of each of the realizations with respect to the reference sentence. The realizations were then scored with a number of automated evaluation metrics developed for machine translation. In order to investigate the correlation of targeted metrics with human evaluations, and gather other acceptable realizations for future evaluations, the judges manually repaired each unacceptable realization during the rating task. In contrast to previous NLG meta-evaluations, we found that several of the metrics correlate moderately well with human judgments of both adequacy and fluency, with the TER family performing best. However, when looking at statistically significant system-level differences in human judgments, we found that some of the metrics get some of the rankings correct, but none get them all correct, with different metrics making different ranking errors. This suggests that multiple metrics should be routinely consulted when comparing realizer systems. Overall, our methodology is similar to that of previous MT meta-evaluations, in that we collected human judgments of system outputs, and compared these scores with those assigned by automatic metrics. A recent alternative approach to paraphrase evaluation is ParaMetric (Callison-Burch et al., 2008); however, it requires a corpus of annotated (aligned) paraphrases (which does not yet exist for PTB data), and is arguably focused more on paraphrase analysis than paraphrase generation. The plan of the paper is as follows: Section 2 discusses the preparation of the corpus of surface realizations. Section 3 describes the human evaluation task and the automated metrics applied. Sections 4 and 5 present and discuss the results of these evaluations. We conclude with some general observations about automatic evaluation of surface realizers, and some directions for further research. 565 2 Data Preparation We collected realizations of the sentences in Section 00 of the WSJ corpus from the following three sources: 1. OpenCCG, a CCG-based chart realizer (White, 2006) 2. The XLE Generator, a LFG-based system developed by Xerox PARC (Crouch et al., 2008) 3. WordNet synonym substitutions, to investigate how differences in lexical choice compare to grammar-based variation.1 Although all three systems used Section 00 of the PTB, they were applied with various parameters (e.g., language models, multiple-output versus single-output) and on different input structures. Accordingly, our study does not compare OpenCCG to XLE, or either of these to the WordNet system. 2.1 OpenCCG realizations OpenCCG is an open source parsing/realization library with multimodal extensions to CCG (Baldridge, 2002). The OpenCCG chart realizer takes logical forms as input and produces strings by combining signs for lexical items. Alternative realizations are scored using integrated n-gram and perceptron models. For robustness, fragments are greedily assembled when necessary. Realizations were generated from 1,895 gold standard logical forms, created by constrained parsing of development-section derivations. The following OpenCCG models (which differ essentially in the way the output is ranked) were used: 1. Baseline 1: Output ranked by a trigram word model 2. Baseline 2: Output ranked using three language models (3-gram words 3-gram words with named entity class replacement factored language model of words, POS tags and CCG supertags) + + 1Not strictly surface realizations, since they do not involve an abstract input specification, but for simplicity we refer to them as realizations throughout. 3. Baseline 3: Perceptron with syntax features and the three LMs mentioned above 4. Perceptron full-model: n-best realizations ranked using perceptron with syntax features and the three n-gram models, as well as discriminative n-grams The perceptron model was trained on sections 0221 of the CCGbank, while a grammar extracted from section 00-21 was used for realization. In addition, oracle supertags were inserted into the chart during realization. The purpose of such a non-blind testing strategy was to evaluate the quality of the output produced by the statistical ranking models in isolation, rather than focusing on grammar coverage, and avoid the problems associated with lexical smoothing, i.e. lexical categories in the development section not being present in the training section. To enrich the variation in the generated realizations, dative-alternation was enforced during realization by ensuring alternate lexical categories of the verb in question, as in the following example: (1) the executives gave [the chefs] [a standing ovation] (2) the executives gave [a standing ovation] [to the chefs] 2.2 XLE realizations The corpus of realizations generated by the XLE system contained 42,527 surface realizations of approximately 1,421 section 00 sentences (an average of 30 per sentence), initially unranked. The LFG f-structures used as input to the XLE generator were derived from automatic parses, as described in (Riezler et al., 2002). The realizations were first tokenized using Penn Treebank conventions, then ranked using perplexities calculated from the same trigram word model used with OpenCCG. For each sentence, the top 4 realizations were selected. The XLE generator provides an interesting point of comparison to OpenCCG as it uses a manuallydeveloped grammar with inputs that are less abstract but potentially noisier, as they are derived from automatic parses rather than gold-standard ones. 566 2.3 WordNet synonymizer To produce an additional source of variation, the nouns and verbs of the sentences in section 00 of the PTB were replaced with all of their WordNet synonyms. Verb forms were generated using verb stems, part-of-speech tags, and the morphg tool.2 These substituted outputs were then filtered using the n-gram data which Google Inc. has made available.3 Those without any 5-gram matches centered on the substituted word (or 3-gram matches, in the case of short sentences) were eliminated. 3 Evaluation From the data sources described in the previous sec- tion, a corpus of realizations to be evaluated by the human judges was constructed by randomly choosing 305 sentences from section 00, then selecting surface realizations of these sentences using the following algorithm: 1. Add OpenCCG’s best-scored realization. 2. Add other OpenCCG realizations until all four models are represented, to a maximum of 4. 3. Add up to 4 realizations from either the XLE system or the WordNet pool, chosen randomly. The intent was to give reasonable coverage of all realizer systems discussed in Section 2 without overloading the human judges. “System” here means any instantiation that emits surface realizations, including various configurations of OpenCCG (using different language models or ranking systems), and these can be multiple-output, such as an n-best list, or single-output (best-only, worst-only, etc.). Accordingly, more realizations were selected from the OpenCCG realizer because 5 different systems were being represented. Realizations were chosen randomly, rather than according to sentence types or other criteria, in order to produce a representative sample of the corpus. In total, 2,114 realizations were selected for evaluation. 2http : //www. informatics . sussex. ac .uk/ re search/ groups / nlp / carro l /morph .html l 3http : //www . ldc . upenn .edu/Catalog/docs/ LDC2 0 0 6T 13 / readme .txt 3.1 Human judgments Two human judges evaluated each surface realization on two criteria: adequacy, which represents the extent to which the output conveys all and only the meaning of the reference sentence; and fluency, the extent to which it is grammatically acceptable. The realizations were presented to the judges in sets containing a reference sentence and the 1-8 outputs selected for that sentence. To aid in the evaluation of adequacy, one sentence each of leading and trailing context were displayed. Judges used the guidelines given in Figure 1, based on the scales developed by the NIST Machine Translation Evaluation Workshop. In addition to rating each realization on the two five-point scales, each judge also repaired each output which he or she did not judge to be fully adequate and fluent. An example is shown in Figure 2. These repairs resulted in new reference sentences for a substantial number of sentences. These repaired realizations were later used to calculate targeted versions of the evaluation metrics, i.e., using the repaired sentence as the reference sentence. Although targeted metrics are not fully automatic, they are of interest because they allow the evaluation algorithm to focus on what is actually wrong with the input, rather than all textual differences. Notably, targeted TER (HTER) has been shown to be more consistent with human judgments than human annotators are with one another (Snover et al., 2006). 3.2 Automatic evaluation The realizations were also evaluated using seven automatic metrics: • IBM’s BLEU, which scores a hypothesis by counting n-gram matches with the reference sentence (Papineni et al., 2001), with smoothing as described in (Lin and Och, 2004) • • • • • • The NIST n-gram evaluation metric, similar to BLEU, but rewarding rarer n-gram matches, and using a different length penalty METEOR, which measures the harmonic mean of unigram precision and recall, with a higher weight for recall (Banerjee and Lavie, 2005) 567 TER (Translation Edit Rate), a measure of the number of edits required to transform a hypothesis sentence into the reference sentence (Snover et al., 2006) TERP, an augmented version of TER which performs phrasal substitutions, stemming, and checks for synonyms, among other improvements (Snover et al., 2009) TERPA, an instantiation of TERP with edit weights optimized for correlation with adequacy in MT evaluations GTM (General Text Matcher), a generaliza- tion of the F-measure that rewards contiguous matching spans (Turian et al., 2003) Additionally, targeted versions of BLEU, METEOR, TER, and GTM were computed by using the human-repaired outputs as the reference set. The human repair was different from the reference sentence in 193 cases (about 9% of the total), and we expected this to result in better scores and correlations with the human judgments overall. 4 Results 4.1 Human judgments Table 1 summarizes the dataset, as well as the mean adequacy and fluency scores garnered from the human evaluation. Overall adequacy and fluency judgments were high (4.16, 3.63) for the realizer systems on average, and the best-rated realizer systems achieved mean fluency scores above 4. 4.2 Inter-annotator agreement Inter-annotator agreement was measured using the κ-coefficient, which is commonly used to measure the extent to which annotators agree in category P(1A−)P−(PE()E), judgment tasks. κ is defined as where P(A) is the observed agreement 1 b−etPw(eEe)n annotators and P(E) is the probability of agreement due to chance (Carletta, 1996). Chance agreement for this data is calculated by the method discussed in Carletta’s squib. However, in previous work in MT meta-evaluation, Callison-Burch et al. (2007), assume the less strict criterion of uniform chance agreement, i.e. for a five-point scale. They also 51 Score Adequacy Fluency 5All the meaning of the referencePerfectly grammatical 4 Most of the meaning Awkward or non-native; punctuation errors 3 Much of the meaning Agreement errors or minor syntactic problems 2 Meaning substantially different Major syntactic problems, such as missing words 1 Meaning completely different Completely ungrammatical Figure Ref. Realiz. Repair 1: Rating scale and guidelines It wasn’t clear how NL and Mr. Simmons would respond if Georgia Gulf spurns them again It weren’t clear how NL and Mr. Simmons would respond if Georgia Gulf again spurns them It wasn’t clear how NL and Mr. Simmons would respond if Georgia Gulf again spurns them Figure 2: Example of repair introduce the notion of “relative” κ, which measures how often two or more judges agreed that A > B, A = B, or A < B for two outputs A and B, irrespective of the specific values given on the five-point scale; here, uniform chance agreement is taken to be We report both absolute and relative κ in Table 2, using actual chance agreement rather than uniform chance agreement. 31. The κ scores of0.60 for adequacy and 0.63 for fluency across the entire dataset represent “substantial” agreement, according to the guidelines discussed in (Landis and Koch, 1977), better than is typically reported for machine translation evaluation tasks; for example, Callison-Burch et al. (2007) reported “fair” agreement, with κ = 0.281 for fluency and κ = 0.307 for adequacy (relative). Assuming the uniform chance agreement that the previously cited work adopts, our inter-annotator agreements (both absolute and relative) are still higher. This is likely due to the generally high quality of the realizations evaluated, leading to easier judgments. 4.3 Correlation with automatic evaluation To determine how well the automatic evaluation methods described in Section 3 correlate with the human judgments, we averaged the human judgments for adequacy and fluency, respectively, for each of the rated realizations, and then computed both Pearson’s correlation coefficient and Spearman’s rank correlation coefficient between these scores and each of the metrics. Spearman’s correlation makes fewer assumptions about the distribu- tion of the data, but may not reflect a linear rela568 tionship that is actually present. Both are frequently reported in the literature. Due to space constraints, we show only Spearman’s correlation, although the TER family scored slightly better on Pearson’s coefficient, relatively. The results for Spearman’s correlation are given in Table 3. Additionally, the average scores for adequacy and fluency were themselves averaged into a single score, following (Snover et al., 2009), and the Spearman’s correlation of each of the automatic metrics with these scores are given in Table 4. All reported correlations are significant at p < 0.001. 4.4 Bootstrap sampling of correlations For each of the sub-corpora shown in Table 1, we computed confidence intervals for the correlations between adequacy and fluency human scores with selected automatic metrics (BLEU, HBLEU, TER, TERP, and HTER) as described in (Koenh, 2004). We sampled each sub-corpus 1000 times with replace- ment, and calculated correlations between the rankings induced by the human scores and those induced by the metrics for each reference sentence. We then used these coefficients to estimate the confidence interval, after excluding the top 25 and bottom 25 coefficients, following (Lin and Och, 2004). The results of this for the BLEU metric are shown in Table 5. We determined which correlations lay within the 95% confidence interval of the best performing metric in each row of Table Table 3; these figures are italicized. 5 Discussion 5.1 Human judgments of systems The results for the four OpenCCG perceptron models mostly confirm those reported in (White and Rajkumar, 2009), with one exception: the B-3 model was below B-2, though the P-B (perceptron-best) model still scored highest. This may have been due to differences in the testing scenario. None of the differences in adequacy scores among the individual systems are significant, with the exception of the WordNet system. In this case, the lack of wordsense disambiguation for the substituted words results in a poor overall adequacy score (e.g., wage floor → wage story). Conversely, it scores highest ffoloro fluency, as substituting a noun or tve srcbo rwesith h a synonym does not usually introduce ungrammaticality. 5.2 Correlations of human judgments with MT metrics Of the non-human-targeted metrics evaluated, BLEU and TER/TERP demonstrate the highest correlations with the human judgments of fluency (r = 0.62, 0.64). The TER family of evaluation metrics have been observed to perform very well in MTevaluation tasks, and although the data evaluated here differs from typical MT data in some important ways, the correlation of TERP with the human judgments is substantial. In contrast with previous MT evaluations where TERP performs considerably better than TER, these scored close to equal on our data, possibly because TERP’s stem, synonym, and paraphrase matching are less useful when most of the variation is syntactic. The correlations with BLEU and METEOR are lower than those reported in (Callison-Burch et al., 2007); in that study, BLEU achieved adequacy and fluency correlations of 0.690 and 0.722, respectively, and METEOR achieved 0.701 and 0.719. The correlations for these metrics might be expected to be lower for our data, since overall quality is higher, making the metrics’ task more difficult as the outputs involve subtler differences between acceptable and unacceptable variation. The human-targeted metrics (represented by the prefixed H in the data tables) correlated even more strongly with the human judgments, compared to the non-targeted versions. HTER demonstrated the best 569 correlation with realizer fluency (r = 0.75). For several kinds of acceptable variation involving the rearrangement of constituents (such as dative shift), TERP gives a more reasonable score than BLEU, due to its ability to directly evaluate phrasal shifts. The following realization was rated 4.5 for fluency, and was more correctly ranked by TERP than BLEU: (3) Ref: The deal also gave Mitsui access to a high-tech medical product. (4) Realiz.: The deal also gave access to a high-tech medical product to Mitsui. For each reference sentence, we compared the ranking of its realizations induced from the human scores to the ranking induced from the TERP score, and counted the rank errors by the latter, informally categorizing them by error type (see Table 7). In the 50 sentences with the highest numbers of rank errors, 17 were affected by punctuation differences, typically involving variation in comma placement. Human fluency judgments of outputs with only punctuation problems were generally high, and many realizations with commas inserted or removed were rated fully fluent by the annotators. However, TERP penalizes such insertions or deletions. Agreement errors are another frequent source of ranking errors for TERP. The human judges tended to harshly penalize sentences with number-agreement or tense errors, whereas TERP applies only a single substitution penalty for each such error. We expect that with suitable optimization of edit weights to avoid over-penalizing punctuation shifts and underpenalizing agreement errors, TERP would exhibit an even stronger correlation with human fluency judgments. None of the evaluation metrics can distinguish an acceptable movement of a word or constituent from an unacceptable movement, with only one reference sentence. A substantial source of error for both TERP and BLEU is variation in adverbial placement, as shown in (7). Similar errors are seen with prepositional phrases and some commonly-occurring temporal adverbs, which typically admit a number of variations in placement. Another important example of acceptable variation which these metrics do not generally rank correctly is dative alternation: Ref. We need to clarify what exactly is wrong with it. Realiz. Flu. TERP BLEU We need to clarify exactly what is wrong with it.50.10.5555 We need to clarify exactly what ’s wrong with it. 5 0.2 0.4046 (7) We need to clarify what , exactly , is wrong with it. 5 0.2 0.5452 We need to clarify what is wrong with it exactly. 4.5 0.1 0.6756 We need to clarify what exactly , is wrong with it. 4 0.1 0.7017 We need to clarify what , exactly is wrong with it. 4 0.1 0.7017 We needs to clarify exactly what is wrong with it. (5) Ref. When test booklets were passed out 48 hours ahead of time, she says she copied questions in the social studies section and gave the answers to students. (6) Realiz. When test booklets were passed out 48 hours ahead of time , she says she copied questions in the social studies section and gave students the answers. The correlations of each of the metrics with the human judgments of fluency for the realizer systems indicate at least a moderate relationship, in contrast with the results reported in (Stent et al., 2005) for paraphrase data, which found an inverse correlation for fluency, and (Cahill, 2009) for the output ofa surface realizer for German, which found only a weak correlation. However, the former study employed a corpus-based paraphrase generation system rather than grammar-driven surface realizers, and the resulting paraphrases exhibited much broader variation. In Cahill’s study, the outputs of the realizer were almost always grammatically correct, and the automated evaluation metrics were ranking markedness instead of grammatical acceptability. 5.3 System-level comparisons In order to investigate the efficacy of the metrics in ranking different realizer systems, or competing realizations from the same system generated using different ranking models, we considered seven different “systems” from the whole dataset of realizations. These consisted of five OpenCCG-based realizations (the best realization from three baseline models, and the best and the worst realization from the full perceptron model), and two XLE-based sys- tems (the best and the worst realization, after ranking the outputs of the XLE realizer with an n-gram model). The mean of the combined adequacy and 570 3 0.103 0.346 fluency scores of each of these seven systems was compared with that of every other system, resulting in 21 pairwise comparisons. Then Tukey’s HSD test was performed to determine the systems which differed significantly in terms of the average adequacy and fluency rating they received.4 The test revealed five pairwise comparisons where the scores were significantly different. Subsequently, for each of these systems, an overall system-level score for each of the MT metrics was calculated. For the five pairwise comparisons where the adequacy-fluency group means differed significantly, we checked whether the metric ranked the systems correctly. Table 8 shows the results of a pairwise comparison between the ranking induced by each evaluation metric, and the ranking induced by the human judgments. Five of the seven non- targeted metrics correctly rank more than half of the systems. NIST, METEOR, and GTM get the most comparisons right, but neither NIST nor GTM correctly rank the OpenCCG-baseline model 1 with respect to the XLE-best model. TER and TERP get two of the five comparisons correct, and they incorrectly rank two of the five OpenCCG model comparisons, as well as the comparison between the XLE-worst and OpenCCG-best systems. For the targeted metrics, HNIST is correct for all five comparisons, while neither HBLEU nor HMETEOR correctly rank all the OpenCCG models. On the other hand, HTER and HGTM incorrectly rank the XLE-best system versus OpenCCG-based models. In summary, some of the metrics get some of the rankings correct, but none of the non-targeted metrics get all of them correct. Moreover, different metrics make different ranking errors. This argues for 4This particular test was chosen since it corrects for multiple post-hoc analyses conducted on the same data-set. the use of multiple metrics in comparing realizer systems. 6 Conclusion Our study suggests that although the task of evaluating the output from realizer systems differs from the task of evaluating machine translations, the automatic metrics used to evaluate MT outputs deliver moderate correlations with combined human fluency and adequacy scores when used on surface realizations. We also found that the MT-evaluation metrics are useful in evaluating different versions of the same realizer system (e.g., the various OpenCCG realization ranking models), and finding cases where a system is performing poorly. As in MT-evaluation tasks, human-targeted metrics have the highest correlations with human judgments overall. These results suggest that the MT-evaluation metrics are useful for developing surface realizers. However, the correlations are lower than those reported for MT data, suggesting that they should be used with caution, especially for cross-system evaluation, where consulting multiple metrics may yield more reliable comparisons. In our study, the targeted version of TERP correlated most strongly with human judgments of fluency. In future work, the performance of the TER family of metrics on this data might be improved by opti- mizing the edit weights used in computing its scores, so as to avoid over-penalizing punctuation movements or under-penalizing agreement errors, both of which were significant sources of ranking errors. Multiple reference sentences may also help mitigate these problems, and the corpus of human-repaired realizations that has resulted from our study is a step in this direction, as it provides multiple references for some cases. We expect the corpus to also prove useful for feature engineering and error analysis in developing better realization models.5 Acknowledgements We thank Aoife Cahill and Tracy King for providing us with the output of the XLE generator. We also thank Chris Callison-Burch and the anonymous reviewers for their helpful comments and suggestions. 5The corpus can be downloaded from http : / /www . l ing .ohio-st ate . edu / ˜mwhite / dat a / emnlp 10 / . 571 This material is based upon work supported by the National Science Foundation under Grant No. 0812297. References Jason Baldridge. 2002. Lexically Specified Derivational Control in Combinatory Categorial Grammar. Ph.D. thesis, University of Edinburgh. S. Banerjee and A. Lavie. 2005. METEOR: An automatic metric for MT evaluation with improved correlation with human judgments. In Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization, pages 65–72. R. Barzilay and L. Lee. 2003. Learning to paraphrase: An unsupervised approach using multiple-sequence alignment. In proceedings of HLT-NAACL, volume 2003, pages 16–23. Aoife Cahill. 2009. Correlating human and automatic evaluation of a german surface realiser. In Proceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 97–100, Suntec, Singapore, August. Association for Computational Linguistics. C. Callison-Burch, M. Osborne, and P. Koehn. 2006. Reevaluating the role of BLEU in machine translation research. In Proceedings of EACL, volume 2006, pages 249–256. Chris Callison-Burch, Cameron Fordyce, Philipp Koehn, Christof Monz, and Josh Schroeder. 2007. (meta-) evaluation ofmachine translation. In StatMT ’07: Proceedings of the Second Workshop on Statistical Machine Translation, pages 136–158, Morristown, NJ, USA. Association for Computational Linguistics. C. Callison-Burch, T. Cohn, and M. Lapata. 2008. Parametric: An automatic evaluation metric for paraphrasing. In Proceedings of the 22nd International Conference on Computational Linguistics-Volume 1, pages 97–104. Association for Computational Linguistics. J. Carletta. 1996. Assessing agreement on classification tasks: the kappa statistic. Computational linguistics, 22(2):249–254. Dick Crouch, Mary Dalrymple, Ron Kaplan, Tracy King, John Maxwell, and Paula Newman. 2008. Xle documentation. Technical report, Palo Alto Research Center. Philip Koenh. 2004. Statistical significance tests for machine translation evaluation. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing. J.R. Landis and G.G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics, 33(1): 159–174. Lin and Franz Josef Och. 2004. Orange: a method for evaluating automatic evaluation metrics for machine translation. In COLING ’04: Proceedings Chin-Yew of the 20th international conference on Computational 501, Morristown, NJ, USA. Associfor Computational Linguistics. Papineni, S. Roukos, T. Ward, and W. Zhu. 2001. Linguistics, page ation K. Bleu: a method for automatic evaluation of machine translation. E. Technical report, IBM Research. Reiter and A. Belz. 2009. An investigation into the validity of some metrics for automatically evaluating natural language generation systems. Computational Linguistics, 35(4):529–558. Stefan Riezler, Tracy H. King, Ronald M. Kaplan, Richard Crouch, John T. III Maxwell, and Mark Johnson. 2002. Parsing the wall street journal using a lexical-functional grammar and discriminative estimation techniques. In Proceedings of 40th Annual Meeting of the Association for Computational Linguistics, pages 271–278, Philadelphia, Pennsylvania, USA, July. Association for Computational Linguistics. Matthew Snover, Bonnie Dorr, Richard Schwartz, Linnea Micciulla, and John Makhoul. 2006. A study of translation edit rate with targeted human annotation. In In Proceedings of Association for Machine Translation in the Americas, pages 223–23 1. M. Snover, N. Madnani, B.J. Dorr, and R. Schwartz. 2009. Fluency, adequacy, or HTER?: exploring different human judgments with a tunable MT metric. In Proceedings of the Fourth Workshop on Statistical Machine Translation, pages 259–268. Association for Computational Linguistics. Amanda Stent, Matthew Marge, and Mohit Singhai. 2005. Evaluating evaluation methods for generation in the presence of variation. In Proceedings of CICLing. J.P. Turian, L. Shen, and I.D. Melamed. 2003. Evaluation of machine translation and its evaluation. recall (C— R), 100:2. Michael White and Rajakrishnan Rajkumar. 2009. Perceptron reranking for CCG realization. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 410–419, Singapore, August. Association for Computational Linguistics. Michael White. 2006. Efficient Realization of Coordinate Structures in Combinatory Categorial Grammar. Research on Language and Computation, 4(1):39–75. 572 Table 1: Descriptive statistics Table 2: Corpora-wise inter-annotator agreement (absolute and relative κ values shown) SXAROWlpeyLos-aErFndAliCzueqrtd-GAFluq0 N.354217690 B.356219470M .35287410G .35241780 TP.465329170T.A34521670T.465230 H.54T76321H0 .543N89270H.653B7491280H.563M41270H.5643G218 Table 3: Spearman’s correlations among NIST (N), BLEU (B), METEOR (M), GTM (G), TERp (TP), TERpa (TA), TER (T), human variants (HN, HB, HM, HT, HG) and human judgments (-Adq: adequacy and -Flu: Fluency); Scores which fall within the 95 %CI of the best are italicized. SROXAWlLeypoasErldniCze rtG0 N.35246 190 B.5618740 M.542719G0 .5341890T .P632180T.A54268 0T .629310 H.7T6 3985H0 .546N180 H.765B8730H.673M5190 H.56G 4318 Table 4: Spearman’s correlations among NIST (N), BLEU (B), METEOR (M), GTM (G), TERp (TP), TERpa (TA), TER (T), human variants (HN, HB, HM, HT, HG) and human judgments (combined adequacy and fluency scores) 573 SRAXOWylLpeosatrEldniezCm rtG0S A.p61d35q94107 .5304%65874L09.5462%136U0SF .lp256u 1209 .51 6%9213L0 .562%91845U Table 5: Spearman’s correlation analysis (bootstrap sampling) of the BLEU scores of various systems with human adequacy and fluency scores SRXOAWylLpeosarEndiCztGH J -12 0 N.6543210 B.6512830 M.4532 960 G.13457960T.P56374210T.A45268730T.562738140 H.7T6854910H.56N482390H.675B1398240H.567M3 240H.56G41290H.8J71562- Table 6: Spearman’s correlations of NIST (N), BLEU (B), METEOR (M), GTM (G), TERp (TP), TERpa (TA), human variants (HT, HN, HB, HM, HG), and individual human judgments (combined adq. and flu. scores) Factor Count Punctuation17 Adverbial shift Agreement Other shifts Conjunct rearrangement Complementizer ins/del PP shift 16 14 8 8 5 4 Table 7: Factors influencing TERP ranking errors for 50 worst-ranked realization groups Table 8: Metric-wise ranking performance in terms of agreement with a ranking induced by combined adequacy and fluency scores; each metric gets a score out of 5 (i.e. number of system-level comparisons that emerged significant as per the Tukey’s HSD test) Legend: Perceptron Best (PB); Perceptron Worst (PW); XLE Best (XB); XLE Worst (XW); OpenCCG baseline models 1 to 3 (C1 ... C3) 574
2 0.95912611 96 emnlp-2010-Self-Training with Products of Latent Variable Grammars
Author: Zhongqiang Huang ; Mary Harper ; Slav Petrov
Abstract: Mary Harper†‡ ‡HLT Center of Excellence Johns Hopkins University Baltimore, MD mharpe r@ umd .edu Slav Petrov∗ ∗Google Research 76 Ninth Avenue New York, NY s lav@ google . com ting the training data and eventually begins over- fitting (Liang et al., 2007). Moreover, EM is a loWe study self-training with products of latent variable grammars in this paper. We show that increasing the quality of the automatically parsed data used for self-training gives higher accuracy self-trained grammars. Our generative self-trained grammars reach F scores of 91.6 on the WSJ test set and surpass even discriminative reranking systems without selftraining. Additionally, we show that multiple self-trained grammars can be combined in a product model to achieve even higher accuracy. The product model is most effective when the individual underlying grammars are most diverse. Combining multiple grammars that were self-trained on disjoint sets of unlabeled data results in a final test accuracy of 92.5% on the WSJ test set and 89.6% on our Broadcast News test set.
3 0.94995946 77 emnlp-2010-Measuring Distributional Similarity in Context
Author: Georgiana Dinu ; Mirella Lapata
Abstract: The computation of meaning similarity as operationalized by vector-based models has found widespread use in many tasks ranging from the acquisition of synonyms and paraphrases to word sense disambiguation and textual entailment. Vector-based models are typically directed at representing words in isolation and thus best suited for measuring similarity out of context. In his paper we propose a probabilistic framework for measuring similarity in context. Central to our approach is the intuition that word meaning is represented as a probability distribution over a set of latent senses and is modulated by context. Experimental results on lexical substitution and word similarity show that our algorithm outperforms previously proposed models.
4 0.9457022 97 emnlp-2010-Simple Type-Level Unsupervised POS Tagging
Author: Yoong Keok Lee ; Aria Haghighi ; Regina Barzilay
Abstract: Part-of-speech (POS) tag distributions are known to exhibit sparsity a word is likely to take a single predominant tag in a corpus. Recent research has demonstrated that incorporating this sparsity constraint improves tagging accuracy. However, in existing systems, this expansion come with a steep increase in model complexity. This paper proposes a simple and effective tagging method that directly models tag sparsity and other distributional properties of valid POS tag assignments. In addition, this formulation results in a dramatic reduction in the number of model parameters thereby, enabling unusually rapid training. Our experiments consistently demonstrate that this model architecture yields substantial performance gains over more complex tagging — counterparts. On several languages, we report performance exceeding that of more complex state-of-the art systems.1
same-paper 5 0.94118625 99 emnlp-2010-Statistical Machine Translation with a Factorized Grammar
Author: Libin Shen ; Bing Zhang ; Spyros Matsoukas ; Jinxi Xu ; Ralph Weischedel
Abstract: In modern machine translation practice, a statistical phrasal or hierarchical translation system usually relies on a huge set of translation rules extracted from bi-lingual training data. This approach not only results in space and efficiency issues, but also suffers from the sparse data problem. In this paper, we propose to use factorized grammars, an idea widely accepted in the field of linguistic grammar construction, to generalize translation rules, so as to solve these two problems. We designed a method to take advantage of the XTAG English Grammar to facilitate the extraction of factorized rules. We experimented on various setups of low-resource language translation, and showed consistent significant improvement in BLEU over state-ofthe-art string-to-dependency baseline systems with 200K words of bi-lingual training data.
6 0.80288577 89 emnlp-2010-PEM: A Paraphrase Evaluation Metric Exploiting Parallel Texts
7 0.78500158 34 emnlp-2010-Crouching Dirichlet, Hidden Markov Model: Unsupervised POS Tagging with Context Local Tag Generation
8 0.78276807 57 emnlp-2010-Hierarchical Phrase-Based Translation Grammars Extracted from Alignment Posterior Probabilities
9 0.76569307 116 emnlp-2010-Using Universal Linguistic Knowledge to Guide Grammar Induction
10 0.76565927 7 emnlp-2010-A Mixture Model with Sharing for Lexical Semantics
11 0.76284468 98 emnlp-2010-Soft Syntactic Constraints for Hierarchical Phrase-Based Translation Using Latent Syntactic Distributions
12 0.7599628 87 emnlp-2010-Nouns are Vectors, Adjectives are Matrices: Representing Adjective-Noun Constructions in Semantic Space
13 0.75824177 75 emnlp-2010-Lessons Learned in Part-of-Speech Tagging of Conversational Speech
14 0.7338261 29 emnlp-2010-Combining Unsupervised and Supervised Alignments for MT: An Empirical Study
15 0.73113567 94 emnlp-2010-SCFG Decoding Without Binarization
16 0.71604544 78 emnlp-2010-Minimum Error Rate Training by Sampling the Translation Lattice
17 0.71448189 81 emnlp-2010-Modeling Perspective Using Adaptor Grammars
18 0.71046644 22 emnlp-2010-Automatic Evaluation of Translation Quality for Distant Language Pairs
19 0.70705152 42 emnlp-2010-Efficient Incremental Decoding for Tree-to-String Translation
20 0.68780118 115 emnlp-2010-Uptraining for Accurate Deterministic Question Parsing