Bitext Dependency Parsing with Bilingual Subtree Constraints
Article Structure
Abstract
This paper proposes a dependency parsing method that uses bilingual constraints to improve the accuracy of parsing bilingual texts (bitexts).
Introduction
Parsing bilingual texts (bitexts) is crucial for training machine translation systems that rely on syntactic structures on either the source side or the target side, or the both (Ding and Palmer, 2005; Nakazawa et al., 2006).
Motivation
In this section, we use an example to show the idea of using the bilingual subtree constraints to improve parsing performance.
Dependency parsing
For dependency parsing, there are two main types of parsing models (Nivre and McDonald, 2008; Nivre and Kubler, 2006): transition-based (Nivre, 2003; Yamada and Matsumoto, 2003) and graph-based (McDonald et al., 2005; Carreras, 2007).
Bilingual subtree constraints
In this section, we propose an approach that uses the bilingual subtree constraints to help parse source sentences that have translations on the target side.
Experiments
All the bilingual data were taken from the translated portion of the Chinese Treebank (CTB) (Xue et al., 2002; Bies et al., 2007), articles 1-325 of CTB, which have English translations with gold-standard parse trees.
Conclusion
We presented an approach using large automatically parsed monolingual data to provide bilingual subtree constraints to improve bitexts parsing.
Topics
subtrees
Appears in 20 sentences as: subtrees (22)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- The subtrees are extracted from large-scale auto-parsed monolingual data on the target side.Page 2, “Introduction”
- We design bilingual subtree features, as described in Section 4, based on the constraints between the source subtrees and the target subtrees that are verified by the subtree list on the target side.Page 3, “Dependency parsing”
- The source subtrees are from the possible dependency relations.Page 3, “Dependency parsing”
- We use large-scale auto-parsed data to obtain subtrees on the target side.Page 3, “Bilingual subtree constraints”
- Then we generate the mapping rules to map the source subtrees onto the extracted target subtrees .Page 3, “Bilingual subtree constraints”
- These features indicate the information of the constraints between bilingual subtrees , that are called bilingual subtree constraints.Page 3, “Bilingual subtree constraints”
- (2009) propose a simple method to extract subtrees from large-scale monolingual data and use them as features to improve monolingual parsing.Page 3, “Bilingual subtree constraints”
- Following their method, we parse large unannotated data with a monolingual parser and obtain a set of subtrees (STt) in the target language.Page 3, “Bilingual subtree constraints”
- We encode the subtrees into string format that is expressed as st 2 w : hid(—w : hid)+1, where w refers to a word in the subtree and hid refers to the word ID of the word’s head (hid=() means that this word is the root of a subtree).Page 3, “Bilingual subtree constraints”
- From the dependency tree of Figure 3, we obtain the subtrees , as shown in Figure 4 and Figure 5.Page 3, “Bilingual subtree constraints”
- After extraction, we obtain a set of subtrees .Page 3, “Bilingual subtree constraints”
See all papers in Proc. ACL 2010 that mention subtrees.
See all papers in Proc. ACL that mention subtrees.
Back to top.
word alignment
Appears in 11 sentences as: word alignment (10) Word alignments (1)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- In our method, a target-side tree fragment that corresponds to a source-side tree fragment is identified via word alignment and mapping rules that are automatically learned.Page 1, “Abstract”
- Basically, a (candidate) dependency subtree in a source-language sentence is mapped to a subtree in the corresponding target-language sentence by using word alignment and mapping rules that are automatically learned.Page 1, “Introduction”
- Suppose that we have an input sentence pair as shown in Figure l, where the source sentence is in English, the target is in Chinese, the dashed undirected links are word alignment links, and the directed links between words indicate that they have a (candidate) dependency relation.Page 2, “Motivation”
- We obtain their corresponding words “I’El(meat)”, “H3 (use)”, and “X¥(fork)” in Chinese Via the word alignment links.Page 2, “Motivation”
- Then we perform word alignment using a word-level aligner (Liang et al., 2006; DeNero and Klein, 2007).Page 5, “Bilingual subtree constraints”
- Figure 8 shows an example of a processed sentence pair that has tree structures on both sides and word alignment links.Page 5, “Bilingual subtree constraints”
- Then through word alignment links, we obtain the corresponding words of the words of 3758.Page 5, “Bilingual subtree constraints”
- We also keep the word alignment information in the target subtree.Page 5, “Bilingual subtree constraints”
- For the target part, we use the word alignment information to represent the target words that have corresponding source words.Page 5, “Bilingual subtree constraints”
- We first check if the added word is in the span of the corresponding words, which can be obtained through word alignment links.Page 7, “Bilingual subtree constraints”
- Word alignments were generated from the Berkeley Aligner (Liang et al., 2006; DeNero and Klein, 2007) trained on a bilingual corpus having approximately 0.8M sentence pairs.Page 7, “Experiments”
See all papers in Proc. ACL 2010 that mention word alignment.
See all papers in Proc. ACL that mention word alignment.
Back to top.
dependency relation
Appears in 10 sentences as: dependency relation (5) dependency relations (5)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- Suppose that we have an input sentence pair as shown in Figure l, where the source sentence is in English, the target is in Chinese, the dashed undirected links are word alignment links, and the directed links between words indicate that they have a (candidate) dependency relation .Page 2, “Motivation”
- By adding “fork”, we have two possible dependency relations , “meat-with-fork” and “ate-with-for ”, to be verified.Page 2, “Motivation”
- The source subtrees are from the possible dependency relations .Page 3, “Dependency parsing”
- At first, we have a possible dependency relation (represented as a source subtree) of words to be verified.Page 6, “Bilingual subtree constraints”
- If yes, we activate a positive feature to encourage the dependency relation .Page 6, “Bilingual subtree constraints”
- ing the dependency relation indicated in the target parts.Page 7, “Bilingual subtree constraints”
- So we activate the feature “3t022YES” to encourage dependency relations among “% $(signed)”, “El/~J(NULL)”, and “IE 3 (project)”.Page 7, “Bilingual subtree constraints”
- If yes, we say feature “2to32YES” to encourage a dependency relation between “ViPage 7, “Bilingual subtree constraints”
- Then they are used to verify the possible dependency relations among source words.Page 7, “Bilingual subtree constraints”
- So the possible dependency relations are verified by the source and target subtrees.Page 7, “Bilingual subtree constraints”
See all papers in Proc. ACL 2010 that mention dependency relation.
See all papers in Proc. ACL that mention dependency relation.
Back to top.
parsing model
Appears in 10 sentences as: parsing model (5) parsing models (5)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- Our method, thus, requires gold standard trees only on the source side of a bilingual corpus in the training phase, unlike the joint parsing model , which requires gold standard trees on the both sides.Page 1, “Abstract”
- Based on the mapping rules, we design a set of features for parsing models .Page 2, “Introduction”
- For dependency parsing, there are two main types of parsing models (Nivre and McDonald, 2008; Nivre and Kubler, 2006): transition-based (Nivre, 2003; Yamada and Matsumoto, 2003) and graph-based (McDonald et al., 2005; Carreras, 2007).Page 2, “Dependency parsing”
- Our approach can be applied to both parsing models .Page 2, “Dependency parsing”
- In this paper, we employ the graph-based MST parsing model proposed by McDonald and PereiraPage 2, “Dependency parsing”
- In the graph-based parsing model , features are represented for all the possible relations on single edges (two words) or adjacent edges (three words).Page 3, “Dependency parsing”
- A set of bilingual features are designed for the parsing model .Page 3, “Dependency parsing”
- Finally, we design the bilingual subtree features based on the mapping rules for the parsing model .Page 3, “Bilingual subtree constraints”
- However, as described in Section 4.3.1, the generated subtrees are verified by looking up list ST): before they are used in the parsing models .Page 5, “Bilingual subtree constraints”
- based parsing models (Nivre, 2003; Yamada and Matsumoto, 2003).Page 9, “Conclusion”
See all papers in Proc. ACL 2010 that mention parsing model.
See all papers in Proc. ACL that mention parsing model.
Back to top.
dependency parsing
Appears in 8 sentences as: dependency parser (1) Dependency parsing (2) dependency parsing (5)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- This paper proposes a dependency parsing method that uses bilingual constraints to improve the accuracy of parsing bilingual texts (bitexts).Page 1, “Abstract”
- This paper proposes a dependency parsing method, which uses the bilingual constraints that we call bilingual subtree constraints and statistics concerning the constraints estimated from large unlabeled monolingual corpora.Page 1, “Introduction”
- The result is used as additional features for the source side dependency parser .Page 1, “Introduction”
- Section 3 introduces the background of dependency parsing .Page 2, “Introduction”
- For dependency parsing , there are two main types of parsing models (Nivre and McDonald, 2008; Nivre and Kubler, 2006): transition-based (Nivre, 2003; Yamada and Matsumoto, 2003) and graph-based (McDonald et al., 2005; Carreras, 2007).Page 2, “Dependency parsing”
- Figure 3 shows an example of dependency parsing .Page 3, “Dependency parsing”
- Table 2: Dependency parsing results of Chinese-source casePage 8, “Experiments”
- Table 3: Dependency parsing results of English-source casePage 8, “Experiments”
See all papers in Proc. ACL 2010 that mention dependency parsing.
See all papers in Proc. ACL that mention dependency parsing.
Back to top.
sentence pairs
Appears in 8 sentences as: sentence pair (3) sentence pairs (5)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- Suppose that we have an input sentence pair as shown in Figure l, where the source sentence is in English, the target is in Chinese, the dashed undirected links are word alignment links, and the directed links between words indicate that they have a (candidate) dependency relation.Page 2, “Motivation”
- To solve the mapping problems, we use a bilingual corpus, which includes sentence pairs , to automatically generate the mapping rules.Page 5, “Bilingual subtree constraints”
- First, the sentence pairs are parsed by monolingual parsers on both sides.Page 5, “Bilingual subtree constraints”
- Figure 8 shows an example of a processed sentence pair that has tree structures on both sides and word alignment links.Page 5, “Bilingual subtree constraints”
- Figure 8: Example of auto-parsed bilingual sentence pairPage 5, “Bilingual subtree constraints”
- From these sentence pairs , we obtain subtree pairs.Page 5, “Bilingual subtree constraints”
- Note that some sentence pairs were removed because they are not one-to-one aligned at the sentence level (Burkett and Klein, 2008; Huang et al., 2009).Page 7, “Experiments”
- Word alignments were generated from the Berkeley Aligner (Liang et al., 2006; DeNero and Klein, 2007) trained on a bilingual corpus having approximately 0.8M sentence pairs .Page 7, “Experiments”
See all papers in Proc. ACL 2010 that mention sentence pairs.
See all papers in Proc. ACL that mention sentence pairs.
Back to top.
POS tags
Appears in 5 sentences as: POS tag (2) POS tagging (1) POS tags (3)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- For the source part, we replace nouns and verbs using their POS tags (coarse grained tags).Page 5, “Bilingual subtree constraints”
- For example, we have the subtree pair: “H %(society):2-ifl €%(fringe):0” and “fringes(W_2):0-of:1-society(W_1):2”, where “of” does not have a corresponding word, the POS tag of “fiéflsocietyY’ is N, and the POS tag of “53 é%(fringe)” is N. The source part of the rule becomes “N22-N20” and the target part becomes “W_2:0-of:1-W_1:2”.Page 5, “Bilingual subtree constraints”
- For Chinese unannotated data, we used the XIN_CMN portion of Chinese Gigaword Version 2.0 (LDC2009T14) (Huang, 2009), which has approximately 311 million words whose segmentation and POS tags are given.Page 7, “Experiments”
- We used the MMA system (Kruengkrai et al., 2009) trained on the training data to perform word segmentation and POS tagging and used the Baseline Parser to parse all the sentences in the data.Page 7, “Experiments”
- The POS tags were assigned by the MXPOST tagger trained on training data.Page 7, “Experiments”
See all papers in Proc. ACL 2010 that mention POS tags.
See all papers in Proc. ACL that mention POS tags.
Back to top.
parsing algorithm
Appears in 4 sentences as: parsing algorithm (4)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- (2006), which is an extension of the projective parsing algorithm of Eisner (1996).Page 3, “Dependency parsing”
- To use richer second-order information, we also implement parent-child-grandchild features (Carreras, 2007) in the MST parsing algorithm .Page 3, “Dependency parsing”
- The parsing algorithm chooses the tree with the highest score in a bottom-up fashion.Page 3, “Dependency parsing”
- Due to the limitations of the parsing algorithm (McDonald and Pereira, 2006; Carreras, 2007), we only use bigram— and trigram-subtrees in our approach.Page 4, “Bilingual subtree constraints”
See all papers in Proc. ACL 2010 that mention parsing algorithm.
See all papers in Proc. ACL that mention parsing algorithm.
Back to top.
graph-based
Appears in 3 sentences as: graph-based (3)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- For dependency parsing, there are two main types of parsing models (Nivre and McDonald, 2008; Nivre and Kubler, 2006): transition-based (Nivre, 2003; Yamada and Matsumoto, 2003) and graph-based (McDonald et al., 2005; Carreras, 2007).Page 2, “Dependency parsing”
- In this paper, we employ the graph-based MST parsing model proposed by McDonald and PereiraPage 2, “Dependency parsing”
- In the graph-based parsing model, features are represented for all the possible relations on single edges (two words) or adjacent edges (three words).Page 3, “Dependency parsing”
See all papers in Proc. ACL 2010 that mention graph-based.
See all papers in Proc. ACL that mention graph-based.
Back to top.
Treebank
Appears in 3 sentences as: Treebank (3)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- Experiments on the translated portion of the Chinese Treebank show that our system outperforms monolingual parsers by 2.93 points for Chinese and 1.64 points for English.Page 1, “Abstract”
- Experiments on the translated portion of the Chinese Treebank (Xue et al., 2002; Bies et al., 2007) show that our system outperforms state-of-the-art monolingual parsers by 2.93 points for Chinese and 1.64 points for English.Page 2, “Introduction”
- All the bilingual data were taken from the translated portion of the Chinese Treebank (CTB) (Xue et al., 2002; Bies et al., 2007), articles 1-325 of CTB, which have English translations with gold-standard parse trees.Page 7, “Experiments”
See all papers in Proc. ACL 2010 that mention Treebank.
See all papers in Proc. ACL that mention Treebank.
Back to top.
UAS
Appears in 3 sentences as: UAS (3)
In Bitext Dependency Parsing with Bilingual Subtree Constraints
- We reported the parser quality by the unlabeled attachment score ( UAS ), i.e., the percentage of tokens (excluding all punctuation tokens) with correct HEADs.Page 7, “Experiments”
- The results showed that the reordering features yielded an improvement of 0.53 and 0.58 points ( UAS ) for the first- and second-order models respectively.Page 8, “Experiments”
- In total, we obtained an absolute improvement of 0.88 points ( UAS ) for the first-order model and 1.36 points for the second-order model by adding all the bilingual subtree features.Page 8, “Experiments”
See all papers in Proc. ACL 2010 that mention UAS.
See all papers in Proc. ACL that mention UAS.
Back to top.
