Domain Adaptation for Statistical Machine Translation Master Defense By Longyue WANG, Vincent MT Group, NLP2CT Lab, FST, UM Supervised by Prof. Lidia S. Chao, Prof. Derek F. Wong 20/08/2014 Research Scope Computational Linguistics Machine Translation Speech Translation Rule-based MT Text Translation Hybrid MT Domain-Specific Domain-Specific Statistical MT MT Figure 1: Our Research Scope [1] [2] [1] Daniel Jurafsky and James Martin (2008) An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition, Second Edition. Prentice Hall. [2] Wikipedia, http://en.wikipedia.org/wiki/Machine_translation. (2/84) Agenda Introduction Proposed Method I: New Criterion Proposed Method II: Combination Proposed Method III: Linguistics Domain-Specific Online Translator Conclusion (3/84) Part I: Introduction (4/84) The First Question WHAT IS STATISTICAL MACHINE TRANSLATION? 5 Statistical Machine Translation Figure 2: Phrase-based SMT Framework SMT translations are generated on the basis of statistical models whose parameters are derived from the analysis of text corpora [3]. Currently, the most successful approach of SMT is phrase-based SMT, where the smallest translation unit is n-gram consecutive words. [3] Peter F. Brown, Vincent J. Della Pietra, Stephen A. Della Pietra, and Robert L. Mercer. 1993. The mathematics of statistical machine translation: Parameter estimation. Computational Linguistics. 19:263–311. (6/84) Statistical Machine Translation Parallel Corpus Monolingual Corpus Figure 2: Phrase-based SMT Framework Corpus is a collection of texts. e.g., IWSLT2012 official corpus. Bilingual corpus is a collection of text paired with translation into another language. Monolingual corpus, in one (mostly are the target side) language. Corpus may come from different genres, topics etc. (7/84) Statistical Machine Translation Word Alignment Translation Table Reordering Model Figure 2: Phrase-based SMT Framework Word alignment can be mined by the help of EM algorithm. Then extract phrase pairs from word alignment to generate translation table. Distance-based reordering model is a penalty of changing position of translated phrases. (8/84) Statistical Machine Translation Language Model Figure 2: Phrase-based SMT Framework Language model assigns a probability to a sequence of words. (n-gram) [4] l 1 pLM ( s ) p ( wi | wii1n 1 ) (1) i 1 [4] F Song and W B Croft (1999). "A General Language Model for Information Retrieval". Research and Development in Information Retrieval. pp. 279–280.. (9/84) Statistical Machine Translation Source Text Searching Translation Candidates Target Text Decoding Figure 2: Phrase-based SMT Framework I e i 1 i 1 ebest arg max e ( fi | ei )d ( starti endi 1 1) PLM (ei | e1...ei 1 ) (2) Decoding function consists of three components: the phrase translation table, which ensure the foreign phrase to match target ones; reordering model, which reorder the phrases appropriately; and language model, which ensure the output to be fluent. (10/84) The Second Question WHAT IS DOMAIN-SPECIFIC SMT SYSTEM? 11 Typical SMT vs. Domain-Specific SMT Typical SMT systems are trained on a large and broad corpus (i.e., general-domain) and deal with texts with ignoring domain. Performance depends heavily upon the quality and quantity of training data. Outputs preserve semantics of the source side but lack morphological and syntactic correctness. Understandable translation quality. BBC News Example [5]. Input: Hollywood actor Jackie Chan has apologised over his son's arrest on drug-related charges, saying he feels "ashamed" and "sad". Google Output: 好萊塢影星成龍已經道歉了他兒子的被捕與毒品有關的指控,說他 感覺“羞恥”和“悲傷”。 [5] Available at http://www.bbc.com/news/world-asia-china-28871698. (BBC News 20 August 2014.) (12/84) 13 Typical SMT vs. Domain-Specific SMT Domain-Specific SMT systems are trained on a small but relative corpus (i.e., in-domain) and deal with texts from one specific domain. Consider relevance between training data and what we want to translate (test data). Outputs preserve semantics of the source side, morphological and syntactic correctness. Publishable quality. Patent Document Example [6] Input: 本发明涉及新的tetramic酸型化合物,它从CCR-5活性复合物中分离出来,在控制 条件下通过将生物纯的微生物培养液(球毛壳霉Kunze SCH 1705 ATCC 74489)发酵来 制备复合物。[5] ICONIC Translator Output: Novel tetramic acid-type compounds isolated from a CCR-5 active complex produced by fermentation under controlled conditions of a biologically pure culture of the microorganism, Chaetomium globosum Kunze SCH 1705, ATCC 74489 ., pharmaceutical compositions containing the compounds. [6] Chinese Patent WO01/74772《受体拮抗剂趋化因子》. (14/84) The Third Question WHAT IS DOMAIN-SPECIFIC TRANSLATION CHALLENGE? 15 Challenge 1 – Ambiguity Multi-meaning may not coincide in bilingual environment. The English word Mouse refers to both animal and electronic device. While in the Chinese side, they are two words. Choosing wrong translation variants is a potential cause for miscomprehension. 1 2 Figure 3: Translation ambiguity example (16/84) Challenge 2 – Language Style News Domain Try to deliver rich information with very economical language. Short and simple-structure sentence make it easy to understand. A lot of abbreviation, date, named entitles. China's Li Duihong won the women's 25-meter sport pistol Olympic gold with a total of 687.9 points early this morning Beijing time. (Guangming Daily, 1996/07/02) 我国女子运动员李对红今天在女子运动手枪决赛中,以 687.9环战胜所有对手,并创造新的奥运记录。(《光明 日报》 1996年7月2日) (17/84) Challenge 2 – Language Style Law Domain Very rigorous even with duplicated terms. Use fewer pronouns, abbreviations etc. to avoid any ambiguity. High frequency words of shall, may, must, be to. Long sentence with long subordinate clauses. When an international treaty that relates to a contract and which the People’s Republic of China has concluded on participated into has provisions of the said treaty shall be applied, but with the exception of clauses to which the People’s Republic of China has declared reservation. 中华人民共和国缔结或者参加的与合同有关的国际条约同中华人民共 和国法律有不同规定的,适用该国际条约的规定。但是,中华人民共和 国声明保留的条款除外。 (18/84) Challenge 3 – Out-Of-Vocabulary Terminology: words or phrases that mainly occur in specific contexts with specific meanings. Variants, increasing, combination etc. 8.36% BHT 91.64% 2,6-二叔丁基 -4-甲基苯酚 Figure 4: Out-of-Vocabulary Example (19/84) Domain Adaptation As SMT is corpus-driven, domain-specificity of training data with respect to the test data is a significant factor that we cannot ignore. There is a mismatch between the domain of available training data and the target domain. Unfortunately, the training resources in specific domains are usually relatively scarce. In such scenarios, various domain adaptation techniques are employed to improve domain-specific translation quality by leveraging general-domain data. (20/84) Domain Adaptation for SMT Domain adaptation can be employed in different SMT components: word-alignment model, language model, translation model and reordering model. [6] [7] Model Figure 5: Domain Adaptation Approaches [6] Hua, Wu, Wang Haifeng, and Liu Zhanyi. "Alignment model adaptation for domain-specific word alignment." Proceedings of the 43rd Annual Meeting on Association for Computational Linguistics. Association for Computational Linguistics, 2005. [7] Koehn, Philipp, and Josh Schroeder. "Experiments in domain adaptation for statistical machine translation." Proceedings of the Second (21/84) Workshop on Statistical Machine Translation. Association for Computational Linguistics, 2007. Domain Adaptation for SMT Various resources can be used for domain adaptation: monolingual corpora, parallel corpora, comparable corpora, dictionaries and dictionary. [8] Resources Figure 5: Domain Adaptation Approaches [8] Wu, Hua, Haifeng Wang, and Chengqing Zong. "Domain adaptation for statistical machine translation with domain dictionary and monolingual corpora." Proceedings of the 22nd International Conference on Computational Linguistics-Volume 1. Association for (22/84) Computational Linguistics, 2008. Domain Adaptation for SMT Considering supervision, domain adaptation approaches can be decided into supervised, semi-supervised and unsupervised. [9] Supervision Figure 5: Domain Adaptation Approaches [9] Snover, Matthew, Bonnie Dorr, and Richard Schwartz. "Language and translation model adaptation using comparable corpora." (23/84) Proceedings of the Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2008. My Thesis Data Selection: solve the ambiguity and language style problems by moving the data distribution of training corpora to target domain. Domain Focused Web-Crawling: reduce the OOVs by mining in-domain dictionary, parallel and monolingual sentences from comparable corpus (web). Figure 6: My Domain Adaptation Approaches (24/84) Part II: Data Selection (25/84) Definition Selecting data suitable for the domain at hand from large general-domain corpora, under the assumption that a general corpus is broad enough to contain sentences that are similar to those that occur in the domain. General-domain Corpus … Law Subtitle Dialog News Novel SMT System Spoken Domain Figure 7: Data Selection Definition (26/84) Framework – TM Adaptation Source Language Target Language Domain Estimation Score Si ,Ti Sim(Vi , M R ) Source Language Target Language We define the set {<Si>, <Ti>, <Si,Ti>} as Vi. MR is an abstract model representing the target domain. Figure 8: My Data Selection Framework (27/84) Framework – TM Adaptation Source Language Target Language Domain Estimation Source Language Target Language Source Language Target Language • Rank sentence pairs according to score. • Select top K% of general-domain data. • K is a tunable threshold. Figure 8: My Data Selection Framework (28/84) Framework – TM Adaptation Source Language Target Language Translation Model (IN) Translation Model (Final) Log-linear /linear Interpolation Domain Estimation Source Language Target Language Translation Model (Pseudo) n p ( x) exp i hi ( x) i 1 Source Language 0 i 1, i 1 Target Language i n pw ( f | e , a ) i p w , i ( f | e , a ) i 0 0 i 1, i 1 i Figure 8: My Data Selection Framework (29/84) Framework – LM Adaptation Target Language Language Model (IN) Target Language Domain Estimation n Log-linear/Linear Interpolation Language Model (Pseudo) Target Language p ( x) exp i hi ( x) i 1 0 i 1, i 1 i n Language Model (Final) p ( s ) i PLM i ( s ) i 1 0 i 1, i 1 i Figure 8: My Data Selection Framework (30/84) Framework – LM Adaptation Figure 8: My Data Selection Framework (31/84) Related Work Vector space model (VSM), which converts sentences into a term-weighted vector and then applies a vector similarity function to measure the domain relevance. The sentence Si is represented as a vector: Si wi1 , wi 2 ,..., win (3) Standard tf-idf weight: Each sentence Si is represented as a vector (wi1, wi2,…, win), and n is the size of the vocabulary. So wij is calculated as follows: wij tf ij log(idf j ) (4) Cosine measure: The similarity between two sentences is then defined as the cosine of the angle between two vectors. cos SGen S IN SGen S IN (5) (32/84) Related Work Perplexity-based model, which employs n-gram in-domain language models to score the perplexity of each sentence in general-domain corpus. Cross-entropy is the average of the negative logarithm of the word probabilities. n 1 H ( p, q) p( wi ) log q( wi ) N i 1 n log q(w ) i 1 i (6) Perplexity pp can be simply transformed with a base b with respect to which the cross-entropy is measured (e.g., bits or nats). (7) pp b H ( p ,q ) Perplexity and cross-entropy are monotonically related. (33/84) Related Work Until now, there are three perplexity-based variants: The first basic one [13]: H I src ( x) (8) The second is called Moore-Lewis [14]: H I src ( x) H O src ( x) (9) which tries to select the sentences that are more similar to indomain but different to out-of-domain. The third is modified Moore-Lewis [15]: H I src ( x) H O src ( x) H I t g t ( x) H O t g t ( x) (10) which considers both source and target language. [13] Jianfeng Gao, Joshua Goodman, Mingjing Li, and Kai-Fu Lee. 2002. Toward a unified approach to statistical language modeling for Chinese. ACM Transactions on Asian Language Information Processing (TALIP). 1:3–33. [14] Robert C. Moore and William Lewis. 2010. Intelligent selection of language model training data. Proceedings of ACL: Short Papers. pp. 220–224. [15] Amittai Axelrod, Xiaodong He, and Jianfeng Gao. 2011. Domain adaptation via pseudo in-domain data selection. In: Proceedings of (34/84) EMNLP. pp. 355–362. Discussion: Grain Level By reviewing their work, I found VSM-based methods can obtain about 1 BLEU point improvement using 60% of general-domain data [10, 11 and 12]. Perplexity-based approaches allow to discard 50% - 99% of the general corpus resulted in an increase of 1.0 - 1.8 BLEU points [13, 14, 15, 16 and 17]. [10] Bing Zhao, Matthias Eck, and Stephan Vogel. 2004. Language model adaptation for statistical machine translation with structured query models. In Proceedings of the 20th International Conference on Computational Linguistics. Association for Computational Linguistics, Geneva, Switzerland. [11] Almut Silja Hildebrand, Matthias Eck, Stephan Vogel, and Alex Waibel. 2005. Adaptation of the translation model for statistical machine translation information retrieval. In 10th Annual Conference of the European Association for Machine Translation (EAMT 2005). Budapest, Hungary. [12] Yajuan Lü, Jin Huang, and Qun Liu. 2007. Improving statistical machine translation performance by training data selection and optimization. Proceedings of EMNLP-CoNLL. pp. 343–350.. [15] Keiji Yasuda and Eiichiro Sumita. 2008. Method for building sentence-aligned corpus from wikipedia. In 2008 AAAI Workshop on Wikipedia and Artificial Intelligence (WikiAI08). [16] George Foster, Cyril Goutte, and Roland Kuhn. 2010. Discriminative instance weighting for domain adaptation in statistical machine translation. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 451–459. Association for Computational Linguistics, Cambridge, Massachusetts. (35/84) Discussion: Grain Level VSM-based similarity is a simple co-occurrence based matching, which only weights single overlapping words. Perplexity-based similarity considers not only the distribution of terms but also the n-gram word collocation. String-difference can comprehensively consider word overlap, n-gram collocation and word position. Edit-distance Word Position Perplexity-based Word Order Query Sentence (V) Cosine tf-idf Word Overlap Figure 9: Data Selection Pyramid Candidate Sentence (R) (36/84) The First Proposed Method EDIT DISTANCE: A NEW DATA SELECTION CRITERION FOR SMT DOMAIN ADAPTATION 37 New Criterion String-difference metric is a better similarity function [21], with higher grain level. Edit-distance is proposed as a new selection criterion. Given a sentence sG from general-domain corpus and a sentence sI from in-domain corpus, the edit distance for these two sequences is defined as the minimum number of edits, i.e. symbol insertions, deletions and substitutions, needed to transform sG into sI. FMS 1 EDword ( sG , sI ) Max( sG , sI ) (11) The normalized similarity score (fuzzy matching score, FMS) is given by Koehn and Senellart [22] in translation memory work. [21] Wang, Longyue, et al. "Edit Distance: A New Data Selection Criterion for Domain Adaptation in SMT." RANLP. 2013. [22] Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, (38/84) Christine Moran et al. 2007. Moses: Open source toolkit for statistical ma-chine translation. Proceedings of ACL. pp. 177–180. New Criterion For each sentence in general-domain corpus, we traverse all in-domain sentences to calculate FMS score and then average them. 1 Score( sG ) N In-domain Corpus N FMS (s G , sI i ) (12) i • • • Figure 10: Edit-distance based data selection General-domain Corpus (39/84) Experiment: Corpora (Chinese-English) General-domain parallel corpus (in-house) includes sentences comparing a various genres such as movie subtitles, law literature, news and novels. In-domain parallel corpus, dev set, test set are randomly selected from the IWSLT2010 Dialog [37], consisting of transcriptions of conversational speech in travel. We use parallel corpora for TM training and the target side for LM training. Data Set Test Set Dev Set In-domain General-domain Sentences 3,500 3,000 17,975 5,211,281 Ave. Len. 9.60 9.46 9.45 12.93 Table 1: Corpora Statistics (English-Chinese) [37] Available at http://iwslt2010.fbk.eu/node/33. (40/84) Experiment: System Setting Baseline: SMT trained on all general-domain corpus; VSM-based system (VSM): SMT trained on top K% of general-domain corpus ranked by Cosine tf-idf metric; Perplexity-based system (PL): SMT trained on top K% of general-domain corpus ranked by basic cross-entropy metric; String-difference system (SD): SMT trained on top K% of general-domain corpus ranked by Edit-distance metric; We investigate K={20, 40, 60, 80}% of ranked generaldomain data as pseudo in-domain corpus for SMT training, where K% means K percentage of general corpus are selected as a subset. (41/84) Experiment: Results Three adaptation methods do better than baseline. VSM can improve nearly 1 BLEU using 80% (more) entire data. PL is a simple but effective method, which increases by 1.1 BLEU using 60% (less) data. SD performs best, which achieve higher BLEU than other two methods with less data. System Baseline VSM PL SD 20% 29.00 (-0.34) 29.45 (+0.11) 29.25 (-0.09) 40% 60% 29.34 29.50 (+0.16) 30.02 (+0.68) 29.65 (+0.31) 30.44 (+1.10) 30.22 (+0.88) 30.97 (+1.63) 80% 30.31 (+0.97) 29.78 (+0.44) 30.21 (+0.87) Table 2: Translation Quality of Adapted Models (42/84) Discussion SD > PL > VSM > Baseline. Higher grained similarity metrics perform better than lower grained ones. Edit-distance Word Position Perplexity-based Word Order Query Sentence (V) Cosine tf-idf Word Overlap Candidate Sentence (R) Figure 9: Data Selection Pyramid However, different grained level methods have their own advanced nature. How about combining the individual models. (43/84) The Second Proposed Method A HYBRID DATA SELECTION MODEL FOR SMT DOMAIN ADAPTATION 44 Combination We investigate the combination of the above three individual models at two levels [23]. Corpus level: weight the pseudo in-domain sub-corpora selected by different methods and then join them together. General-domain Corpus Combined Corpus Figure 11: Combination Approach VSM • • • General-domain Corpus ED [23] Wang, Longyue, et al. "iCPE: A Hybrid Data Selection Model for SMT Domain Adaptation." Chinese Computational Linguistics and (45/84) Natural Language Processing Based on Naturally Annotated Big Data. Springer Berlin Heidelberg, 2013. 280-290. Combination Model level: perform linear interpolation on the translation models trained on difference sub-corpora. n ( f | e ) ii ( f | e ) (13) i 0 n pw ( f | e , a ) i p w , i ( f | e , a ) (14) i 0 where i = 1, 2, and 3 denoting the phrase translation probability and lexical weights trained on the VSM, perplexity and edit-distance’s subsets. αi and βi are the tunable interpolation parameters, subject to i i 1 (46/84) Experiment: Corpora (Chinese-English) General-domain parallel corpus includes sentences comparing a various genres such as movie subtitles, law literature, news and novels etc. Domain News Novel Law Others Total Sent. No. 279,962 304,932 48,754 504,396 1,138,044 % 24.60% 26.79% 4.28% 44.33% 100.00% Table 3: Translation Quality of Adapted Models In-domain parallel corpus, dev set, test set are disjoinedly and randomly selected from LDC corpus [38] (Hong Kong law domain). [38] LDC2004T08, https://catalog.ldc.upenn.edu/LDC2004T08. (47/84) Experiment: Corpora (Chinese-English) Data Set Test Set Dev Set In-domain Training Set Lang. EN ZH EN ZH EN ZH EN ZH Sentences 2,050 2,000 45,621 1,138,044 Tokens 60,399 59,628 59,924 59,054 1,330,464 1,321,655 28,626,367 28,239,747 Av. Len. 29.46 29.09 29.92 29.53 29.16 28.97 25.15 24.81 Table 4: Corpora Statistics Corpus size, data-type distribution, in/gen domain ratio are different. Data selection performance may be different. We use parallel corpora for TM training and the target side for LM training. (48/84) Experiment: System Setting Baseline: the general-domain baseline (GC-Baseline) are respectively trained on entire general corpus. Individual Model: Cosine tf-idf (Cos), proposed editdistance based (ED) and three perplexity-based variants: cross-entropy (CE), Moore-Lewis (ML) and modified MooreLewis (MML). Combined Model: combined Cos, ED and the best perplexity-based model at corpus level (iCPE-C) and model level (named iCPE-M). We report selected corpora in a step of 2x starting from using 3.75% of general corpus K={3.75, 7.5, 15, 30, 60}%. (49/84) Experiment: Individual Model Results Perplexity-based variants are all effective methods. MML performs best: improve highest (nearly 2 BLEU) with least data (15%). MML> ED > CE > ML > Cos > Baseline System GC-Baseline CE ML MML Cos ED 3.75% 37.10 (-) 38.07 (-) 38.26(-) 37.87 (-) 37.70 (-) 7.5% 15% 30% 60% 39.15 39.82 (+0.67) 40.39 (+1.24) 40.79 (+1.64) 39.43(+0.28) 40.33 (+1.18) 40.08 (+0.93) 40.46 (+1.31) 40.27 (+1.12) 40.91 (+1.76) 41.12 (+1.97) 40.02 (+0.87) 39.82 (+0.67) 38.44 (-) 39.45 (+0.30) 40.17 (+1.02) 39.88 (+0.73) 39.00 (-) 40.88 (+1.73) 40.24 (+1.09) 40.00 (+0.85) Table 5: Translation Quality of Adapted Models (50/84) Experiment: Results Good performances are at K={7.5, 15, 30}%, thus we conduct combination methods in this section. Considering different nature of them, we will further combine MML (best perplexity-based), Cos and ED. CE ML MML Cos ED GC-Base 41 40 BLEU 39 38 37 0 20 40 60 80 100 Size of Selected Data K% Figure 12: Combination Approach (51/84) Experiment: Combination Model Results Two combination methods perform better than the best individual model. (slightly) Model-level combination is better than corpus-level one. (+0.23 BLEU) Combination models > individual models > Baseline System 7.5% 15% 30% GC-Baseline 39.15 MML 40.91 (+1.76) 41.12 (+1.97) 40.02 (+0.87) iCPE-C 41.01 (+1.86) 41.95 (+2.80) 41.98 (+2.83) iCPE-M 41.13 (+1.98) 42.21 (+3.06) 41.84 (+2.69) Table 6: Translation Quality of Adapted Models (52/84) Discussion We compare many data selection methods: VSM-based: cosine tf-idf. Perplexity-based: basic cross-entropy, Moore-Lewis and modified Moore-Lewis. String-difference: edit-distance. Combination: Corpus-level and Model-level Above methods only consider word itself (surface information). Languages have a larger set of different words leads to sparsity problems. Weak at capturing language style, sentence structure, sematic information. (53/84) The Third Proposed Method LINGUISTICALLY-AUGMENTED DATA SELECTION FOR SMT DOMAIN ADAPTATION 54 Linguistic DS We explore two more linguistic information for data selection approach [25]: Surface form (f), word itself, have rich lexicon information. Named Entity categories (n) group together proper nouns that belong to the same semantic class (person, location, organization) [26]. Part-Of-Speech tags (t) group together words that share the same grammatical function (e.g. adjectives, nouns, verbs) [27]. [25] Antonio Toral, Pavel Pecina, Longyue Wang, Josef van Genabith. (2014). “Linguistically-augmented Perplexity-based Data Selection for Language Models.” Computer Speech and Language, (accepted and in minor revisions).. [26] E. W. D. Whittaker, P. C. Woodland, Comparison of language modelling techniques for russian and english, in: ICSLP, ISCA, 1998. [27] P. A. Heeman, Pos tags and decision trees for language modeling, in: 1999 Joint SIGDAT Conference on Empirical Methods in Natural (55/84) Language Processing and Very Large Corpora, 1999, pp. 129{137. Linguistic DS Change the original corpus (f) into linguistic format (fn, ft and t) and use them for LM training and sentence scoring. The core metric is the modified Moore-Lewis. According to the scores, select data from original corpus (surface) to train adapted SMT models. H I src ( x) H O src ( x) H I t g t ( x) H O t g t ( x) Need 4 LM models: 1, in-domain corpus in source language 2, in-domain corpus in target language 3, out-of-domain corpus in source language 4, out-of-domain corpus in target language Figure 13: Linguistically-based Data Selection Method (56/84) Linguistic-based DS Based on individual models, we further combine different types of linguistic knowledge: Corpus level: given the sentences selected by all the individual models considered for a given threshold, we traverse the first ranked sentence by each of the methods, then we proceed to the set of second best ranked sentences, and so forth. Model level: Similar. The traversed sentences are kept in different sets. Build LMs on each set and then interpolate them. They are same as the second experiment. (57/84) Experiment: Corpora (Chinese-English) General-domain parallel corpus combined with generaldomain corpora: CWMT2013 [39], UMCorpus [40], News Magazine [41] etc. In-domain parallel corpus, dev set, test set are the IWSLT2014 TED Talk (talk domain) [42]. Data Set (EN/ZH) Test Set Dev Set In-domain General-domain Sentences 1,570 887 177,477 10,021,162 Ave. Len. 26.54/23.41 26.47/23.24 26.47/23.58 23.02/21.36 Table 7: Corpora Statistics [39] http://www.liip.cn/cwmt2013/. [40] http://nlp2ct.cis.umac.mo/um-corpus/. [41] LDC2005T10. https://catalog.ldc.upenn.edu/LDC2005T10. [42] http://workshop2014.iwslt.org/. (58/84) Experiment: System Setting All adapted systems are log-linearly interpolated with the indomain model to further improve performance. Baseline: GI-Baseline is trained on all in-domain corpus and general corpus. Individual Model: surface form based (f), POS based (t), surface+named entity based (fn), surface+POS (ft) . Combined Model: corpus level (Comb-C) and model level (Comb-M). We investigate K={25, 50, 75}% of ranked general-domain data as pseudo in-domain corpus for SMT training. (59/84) Experiment: Individual Model Results After adding more linguistic information, fn and ft can improve baseline by about 1 BLEU. t (only POS) perform poorly due to lack of lexicon information. Considering their performance, we will combine f, fn and ft. System GI-Baseline f 25% 31.91 (-8.29) 50% 40.20 38.83 (-1.37) 75% 41.37 (+1.17) t 21.20 (-19.00) 27.90 (-12.30) 27.90 (-12.30) fn 31.93 (-8.27) 37.86 (-2.34) 40.93 (+0.73) ft 30.00 (-10.20) 38.74 (-1.46) 41.81 (+1.61) Table 8: Translation Quality of Adapted Models (60/84) Experiment: Combination Model Results Both combination methods are better than best individual model (from +0.64 to +0.11 BLEU). Combination may success the advantages of each linguisticbased methods. (lexicon, spacity, language style) High-inflected languages such as English-German may have better performance with more linguistic information. System GI-Baseline f 25% 31.91 (-8.29) 50% 40.20 38.83 (-1.37) 41.37 (+1.17) ft 30.00 (-10.20) 38.74 (-1.46) 41.81 (+1.61) Comb-C 33.01 (-7.19) 39.07 (-1.13) 41.92 (+1.72) Comb-M 32.74 (-7.46) 38.95 (-1.25) 42.01 (+1.81) Table 9: Translation Quality of Adapted Models 75% (61/84) Part III: Real-Life System (62/84) Real-life Environment To prove the robustness and language-independence of some domain adaptation approaches, we evaluation it in reallife system. WMT (since 2005) is most famous workshop with high-quality shared task on machine translation. We attended WMT2014 medical translation task [43]: Czech-English, French-English, German-English. (6 pairs) Very large resources: up to 36 million general-domain parallel sentences and 4 million in-domain parallel sentences. Medical texts are more complex. Chemical formulae, e.g “CH2-(OCH2CH2)n-”. [43] http://www.statmt.org/wmt14/. (63/84) WMT2014 Medical Translation Task By observing the text of medical text, we present a number of detailed domain adaptation techniques and approaches: Task Oriented Pre-processing. Language Model Adaptation. Translation Model Adaptation. Numeric Adaptation. Hyphenated Word Adaptation. Combination above all methods. Figure 14: Results and Rankings of Our System Finally, 1st rank on three language pairs, and 2nd rank on others. (64/84) BenTu System Based these models (medical domain), we develop my first online translator, BenTu, which is a domain-specific multi-tire SMT system [44]. Three layers: pre-processing, decoder and post-processing Easy to add new language pairs and domains Figure 15: Framework of BenTu System [44] The architecture is designed referring to PluTO project: Tinsley, John, Andy Way, and Paraic Sheridan. "PLuTO: MT for online patent (65/84) translation." Association for Machine Translation in the Americas, 2010.. BenTu System Figure 16: User Interface of BenTu System (66/84) Part V: Conclusion (67/84) Thesis Contribution To solve the problems in domain-specific SMT, we proposed Data Selection methods as described. o New data selection criterion o Combination model o Linguistically-augmented data selection Domain Focused Web-Crawling o Integrated models for cross-language document alignment o Combining topic classifier and perplexity for filtering Real-life domain-specific SMT based on a number of adapted models are developed. (68/84) Total Contribution Figure 17: My work in the past three years (69/84) Future Work Data Selection o Graphical model and label propagation o Neural language model Domain Focused Web-Crawling o Improve the performance by mining the in-domain dictionary. Real-life domain-specific SMT o Extend to more language pairs: Chinese, Japanese etc. o Extend to more domains: science technology, laws and news (70/84) My Publications Journal Papers 1, Antonio Toral, Pavel Pecina, Longyue Wang, Josef van Genabith. 2014. Linguistically-augmented Perplexity-based Data Selection for Language Models. Computer Speech and Language (accepted). (IF=1.463) 2, Longyue Wang, Derek F. Wong, Lidia S. Chao, Yi Lu, and Junwen Xing. 2013. A Systematic Comparison of Data Selection Criteria for SMT Domain Adaptation. The Scientific World Journal, vol. 2014, Article ID 745485, 10 pages. (IF=1.730) 3, Long-Yue WANG, Derek F. WONG, Lidia S. CHAO. 2012. TQDL: Integrated Models for Cross-Language Document Retrieval. International Journal of Computational Linguistics and Chinese Language Processing (IJCLCLP), pages 15-32. (THCI Core) Conference Papers 4, Longyue Wang, Yi Lu, Derek F. Wong, Lidia S. Chao, Yiming Wang, Francisco Oliveira. 2014. Combining Domain Adaptation Approaches for Medical Text Translation. In Proceedings of the Ninth Workshop on Statistical Machine Translation. (ACL Anthology and EI) (71/84) My Publications 5, Yi Lu, Longyue Wang, Derek F. Wong, Lidia S. Chao, Yiming Wang, Francisco Oliveira. (2014) "Domain Adaptation for Medical Text Translation using Web Resources". In Proceedings of the Ninth Workshop on Statistical Machine Translation. (ACL Anthology and EI) 6, Yiming Wang, Longyue Wang, Xiaodong Zeng, Derek F. Wong, Lidia S.Chao, Yi Lu. 2014. Factored Statistical Machine Translation for Grammatical Error Correction”, In Proceedings of the Eighth Conference on Computational Natural Language Learning (CoNLL 2014), pages 83-90. (ACL Anthology and EI) 7, Longyue Wang, Derek F. Wong, Lidia S. Chao, Junwen Xing, Yi Lu, Isabel Trancoso. 2013. Edit Distance: A New Data Selection Criterion for SMT Domain Adaptation. In Proceedings of Recent Advances in Natural Language Processing, pages 727-732. (ACL Anthology and EI) 8, Longyue Wang, Derek F. Wong, Lidia S. Chao, Yi Lu, Junwen Xing. 2013. iCPE: A Hybrid Data Selection Model for SMT Domain Adaptation. In Proceedings of the 12th China National Conference on Computational Linguistics (12th CCL), Lecture Notes in Artificial Intelligence (LNAI) Springer series, pages 280-290. (EI) (72/84) My Publications 9, Junwen Xing, Longyue Wang, Derek F. Wong, Lidia S. Chao, Xiaodong Zeng. 2013. UMChecker: A Hybrid System for English Grammatical Error Correction. In Proceedings of the Seventeenth Conference on Computational Natural Language Learning (CoNLL 2013), pages 34-42. (ACL Anthology and EI) 10, Longyue WANG, Shuo Li, Derek F. WONG, Lidia S. CHAO. 2012. A Joint Chinese Named Entity Recognition and Disambiguation System. In Proceeding of the 2th CIPS-SIGHAN Joint Conference on Chinese Language Processing (CLP2012), pages 146-151. (ACL Anthology) 11, Longyue WANG, Derek F. WONG, Lidia S. CHAO, Junwen Xing. 2012. CRFs-Based Chinese Word Segmentation for Micro-Blog with Small-Scale Data. In Proceedings of the Second CIPSSIGHAN Joint Conference on Chinese Language Processing (CLP2012), pages 51-57. (ACL Anthology) 12, Long-Yue Wang, Derek F. WONG, Lidia S. CHAO. 2012. An Experimental Platform for Cross-Language Document Retrieval. The 2012 International Conference on Applied Science and Engineering (ICASE2012), pages 33253329. (EI) (73/84) My Publications 13, Longyue Wang, Derek F. WONG, Lidia S. CHAO. 2012. An Improvement in Cross-Language Document Retrieval Based on Statistical Models. The TwentyFourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012), pages 144-155. (ACL Anthology and EI) 14, Liang Tian, Derek F. Wong, Lidia S. Chao, Paulo Quaresma, Francisco Oliveira, Yi Lu, Shuo Li, Yiming Wang, Longyue Wang. 2014. UM-Corpus: A Large English-Chinese Parallel Corpus for Statistical Machine Translation. In Proceedings of the 9th Edition of its Language Resources and Evaluation Conference (LREC2014), pages 1837-1842. (EI) (74/84) Thank You! 謝謝! Obrigado! (75/84) (76/84) Appendix (77/84) Related Work Zhao et al. [10] firstly use this information retrieval techniques to retrieve sentences from monolingual corpus to build a LM, and then interpolate it with generalbackground LM. Hildebrand et al. [11] extend it to sentence pairs, which are used to train a domain-specific TM. Lü et al. [12] further proposed re-sampling and reweighting methods for online and offline TM optimization. [10] Bing Zhao, Matthias Eck, and Stephan Vogel. 2004. Language model adaptation for statistical machine translation with structured query models. In Proceedings of the 20th International Conference on Computational Linguistics. Association for Computational Linguistics, Geneva, Switzerland. [11] Almut Silja Hildebrand, Matthias Eck, Stephan Vogel, and Alex Waibel. 2005. Adaptation of the translation model for statistical machine translation information retrieval. In 10th Annual Conference of the European Association for Machine Translation (EAMT 2005). Budapest, Hungary. [12] Yajuan Lü, Jin Huang, and Qun Liu. 2007. Improving statistical machine translation performance by training data selection and (78/84) optimization. Proceedings of EMNLP-CoNLL. pp. 343–350.. Related Work In language modeling, Gao et al. [13], Moore and Lewis [14] have used perplexity-based scores adapt LMs. Then it was firstly applied for SMT adaptation by Yasuda et al. [15] and Foster et al. [16]. Axelrod et al. [17] further improve the performance of TM adaptation by considering bilingual information. [13] Jianfeng Gao, Joshua Goodman, Mingjing Li, and Kai-Fu Lee. 2002. Toward a unified approach to statistical language modeling for Chinese. ACM Transactions on Asian Language Information Processing (TALIP). 1:3–33. [14] Robert C. Moore and William Lewis. 2010. Intelligent selection of language model training data. Proceedings of ACL: Short Papers. pp. 220–224. [15] Keiji Yasuda and Eiichiro Sumita. 2008. Method for building sentence-aligned corpus from wikipedia. In 2008 AAAI Workshop on Wikipedia and Artificial Intelligence (WikiAI08). [16] George Foster, Cyril Goutte, and Roland Kuhn. 2010. Discriminative instance weighting for domain adaptation in statistical machine translation. In Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 451–459. Association for Computational Linguistics, Cambridge, Massachusetts. [17] Amittai Axelrod, Xiaodong He, and Jianfeng Gao. 2011. Domain adaptation via pseudo in-domain data selection. In: Proceedings of (79/84) EMNLP. pp. 355–362. Related Work After selection, we obtain pseudo in-domain sub-corpus and in-domain one is available, mixture-modeling is to integrate different language models or translation models. Foster and Kuhn [18] investigate linear and log-linear interpolation for individual language models trained by different corpora. Linear interpolation for SMT has been used a lot [19]. Alternatively, the translation models can be added to the global log-linear SMT model as features, with weights optimized through minimum-error-rate training (MERT) [20]. [18] George Foster and Roland Kuhn. 2007. Mixture-model adaptation for SMT. In Proceedings of the Second Workshop on Statistical Machine Translation, StatMT ’07, pages 128–135. Association for Computational Linguistics, Prague, Czech Republic. [19] Graeme Blackwood, Adrià de Gispert, Jamie Brunning, and William Byrne. 2008. European language translation with weighted finite state transducers: The CUED MT system for the 2008 ACL workshop on SMT. In Proceedings of the Third Workshop on Statistical Machine Translation, pages 131–134. Association for Computational Linguistics, Columbus, Ohio. [20]Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, (80/84) Christine Moran et al. 2007. Moses: Open source toolkit for statistical ma-chine translation. Proceedings of ACL. pp. 177–180. Experimental Setup Overall Running Time: The environment is HPC Cluster Pearl. Computing Node CPU Intel Xeon X5675, 24 cores, 180 GB. Data Selection: Method VSM (GPU) Perplexity String-Diff. (GPU) 2.5 million 5 million 7.5 million 10 million 8 hr 15 hr 29 hr 41 hr 20 min 25 min 30 min 40 min 22 hr 40 hr 62 hr 70 hr 2.5 million 4 hr 1 hr 5 million 13 hr 2 hr 7.5 million 23 hr 4 hr 10 million 32 hr 6 hr SMT: Task Training Tuning (81/84) Experimental Setup Corpus Processing: Propose better data processing steps [29] for domain adaptation task. For Chinese segmentation, we use in-house system [30]. For other languages, we use European tokenizer [31]. Linguistic information are extracted by Stanford CoreNLP toolkits [32]. Others such as case-processing (truecase), length-cleaning (1-80) ect., we use Moses scripts. [29] Longyue Wang, Yi Lu, Derek F. Wong, Lidia S. Chao, Yiming Wang, Francisco Oliveira. (2014) "Combining Domain Adaptation Approaches for Medical Text Translation". In Proceedings of the Ninth Workshop on Statistical Machine Translation. [30] Longyue WANG, Derek F. WONG, Lidia S. CHAO, Junwen Xing. (2012). "CRFs-Based Chinese Word Segmentation for Micro-Blog with Small-Scale Data." Proceedings of the Second CIPS-SIGHAN Joint Conference on Chinese Language Processing (CLP2012), pages 51–57. [31] Philipp Koehn. 2005. Europarl: A parallel corpus for statistical machine translation. MT Summit. Vol. 5. pp. 79–86. [32] Manning, Christopher D., Surdeanu, Mihai, Bauer, John, Finkel, Jenny, Bethard, Steven J., and McClosky, David. 2014. The Stanford CoreNLP Natural Language Processing Toolkit. In Proceedings of 52nd Annual Meeting of the Association for Computational Linguistics: (82/84) System Demonstrations, pp. 55-60 Experimental Setup SMT: Moses decoder [33], a state-of-the-art open-source phrase-based SMT system. The translation and the re-ordering model relied on “growdiag-final” symmetrized word-to-word alignments built using GIZA++ [34]. A 5-gram language model was trained using the IRSTLM toolkit [35], exploiting improved modified Kneser-Ney smoothing, and quantizing both probabilities and back-off weights. [33] Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran et al. 2007. Moses: Open source toolkit for statistical ma-chine translation. Proceedings of ACL. pp. 177–180. [34] Franz Josef Och and Hermann Ney. 2003. A systematic comparison of various statistical alignment models. Computational Linguistics. 29:19–51. [35] Marcello Federico, Nicola Bertoldi, and Mauro Cettolo. 2008. IRSTLM: an open source toolkit for handling large scale language models. (83/84) Proceedings of Inter-speech. pp. 1618–1621. Experimental Setup Data Selection: For Cosine tf-idf and Edit-distance, we develop them on GPU. For Perplexity-based methods, we perform SRILM toolkit [36] to conduct 5-gram LMs with interpolated modified Kneser-Ney discounting. We use end-to-end evaluation method: using BLEU [37] as an evaluation metric to reflect the domain-specific translation quality. [36] Andreas Stolcke and others. 2002. SRILM-an extensible language modeling toolkit. Proceedings of the International Conference on Spoken Language Processing. pp. 901–904. [37] Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic eval-uation of machine translation. (84/84) Proceedings of ACL. pp. 311–318.
© Copyright 2024