Creating manually-annotated syntactic treebanks is an expensive and time consuming task. Recently there has been a great deal of interest in cross-lingual syntactic transfer, where a parsing model is trained for some language of interest, using only treebanks in other languages. There is a clear motivation for this in building parsing models for languages for which treebank data is unavailable. Methods for syntactic transfer include annotation projection methods [Hwa et al.2005, Ganchev et al.2009, McDonald et al.2011, Ma and Xia2014, Rasooli and Collins2015, Lacroix et al.2016, Agić et al.2016], learning of delexicalized models on universal treebanks [Zeman and Resnik2008, McDonald et al.2011, Täckström et al.2013, Rosa and Zabokrtsky2015], treebank translation [Tiedemann et al.2014, Tiedemann2015, Tiedemann and Agić2016] and methods that leverage cross-lingual representations of word clusters, embeddings or dictionaries [Täckström et al.2012, Durrett et al.2012, Duong et al.2015a, Zhang and Barzilay2015, Xiao and Guo2015, Guo et al.2015, Guo et al.2016, Ammar et al.2016].
This paper considers the problem of cross-lingual syntactic transfer with limited resources of monolingual and translation data. Specifically, we use the Bible corpus of christodouloupoulos2014massively as a source of translation data, and Wikipedia as a source of monolingual data. We deliberately limit ourselves to the use of Bible translation data because it is available for a very broad set of languages: the data from christodouloupoulos2014massively includes data from 100 languages. The Bible data contains a much smaller set of sentences (around 24,000) than other translation corpora, for example Europarl [Koehn2005], which has around 2 million sentences per language pair. This makes it a considerably more challenging corpus to work with. Similarly, our choice of Wikipedia as the source of monolingual data is motivated by the availability of Wikipedia data in a very broad set of languages.
We introduce a set of simple but effective methods for syntactic transfer, as follows:
We describe a method for deriving cross-lingual clusters, where words from different languages with a similar syntactic or semantic role are grouped in the same cluster. These clusters can then be used as features in a shift-reduce dependency parser.
We describe a method for transfer of lexical information from the target language into source language treebanks, using word-to-word translation dictionaries derived from parallel corpora. Lexical features from the target language can then be integrated in parsing.
We describe a method that integrates the above two approaches with the density-driven approach to annotation projection described in [Rasooli and Collins2015].
Experiments show that our model outperforms previous work on a set of European languages from the Google universal treebank [McDonald et al.2013]: we achieve 80.9% average unlabeled attachment score (UAS) on these languages; in comparison the work of yuanregina15, guo2016representation and ammar2016one have UAS of 75.4%, 76.3% and 77.8% respectively. All of these previous works make use of the much larger Europarl [Koehn2005] corpus to derive lexical representations. When using Europarl data instead of the Bible, our approach gives 83.9% accuracy, a 1.7% absolute improvements over [Rasooli and Collins2015]. Finally, we conduct experiments on 38 datasets (26 languages) in the universal dependencies v1.3 [Nivre et al.2016] corpus. Our method has an average unlabeled dependency accuracy of 74.8% for these languages, more than 6% higher than the method of rasooli-collins:2015:EMNLP. 13 datasets (10 languages) have accuracies higher than 80.0%.111 The parser code is available at https://github.com/rasoolims/YaraParser/tree/transfer.
This section gives a description of the underlying parsing models used in our experiments, the data sets used, and a baseline approach based on delexicalized parsing models.
2.1 The Parsing Model
We assume that the parsing model is a discriminative linear model, where given a sentence , and a set of candidate parses , the output from the model is
is a parameter vector, andis a feature vector for the pair . In our experiments we use the shift-reduce dependency parser of rasooli2015yara, which is an extension of the approach in [Zhang and Nivre2011]
. The parser is trained using the averaged structured perceptron[Collins2002].
We assume that the feature vector is the concatenation of three feature vectors:
is an unlexicalized set of features. Each such feature may depend on the part-of-speech (POS) tag of words in the sentence, but does not depend on the identity of individual words in the sentence.
is a set of cluster features. These features require access to some dictionary that maps each word in the sentence to an underlying cluster identity. Clusters may for example be learned using the Brown clustering algorithm [Brown et al.1992]. The features may make use of cluster identities in combination with POS tags.
is a set of lexicalized features. Each such feature may depend directly on word identities in the sentence. These features may also depend on part-of-speech tags or cluster information, in conjunction with lexical information.
Appendix A has a full description of the features used in our experiments.
2.2 Data Assumptions
Throughout this paper we will assume that we have source languages , and a single target language . We assume the following data sources:
Source language treebanks.
We have a treebank for each language .
Part-of-speech (POS) data.
We have hand-annotated POS data for all languages . We assume that the data uses a universal POS set that is common across all languages.
We have monolingual raw data for each of the languages. We use to refer to the monolingual data for the ’th language.
We have translation data for all language pairs. We use to refer to translation data for the language pair where and .
In our main experiments we use the Google universal treebank [McDonald et al.2013] as our source language treebanks222We also train our best performing model on the newly released universal treebank v1.3 [Nivre et al.2016]. See §4.3 for more details. (this treebank provides universal dependency relations and POS tags), Wikipedia data as our monolingual data, and the Bible data from christodouloupoulos2014massively as the source of our translation data. In additional experiments we use the Europarl corpus as a source of translation data, in order to measure the impact of using the smaller Bible corpus.
2.3 A Baseline Approach: Delexicalized Parsers with Self-Training
Given the data assumption of a universal POS set, the feature vectors can be shared across languages. A simple approach is then to simply train a delexicalized parser using treebanks , using the representation (see [McDonald et al.2013, Täckström et al.2013]).
Our baseline approach makes use of a delexicalized parser, with two refinements:
We use the six properties from the world atlas of language structures (WALS) [Dryer and Haspelmath2013] to select a subset of closely related languages for each target language. These properties are shown in Table 1. The model for a target language is trained on treebank data from languages where at least 4 out of 6 WALS properties are common between the source and target language.333There was no effort to optimize this choice; future work may consider more sophisticated sharing schemes. This gives a slightly stronger baseline: our experiments showed an improvement in average labeled dependency accuracy for the languages from 62.52% to 63.18%. Table 2 shows the set of source languages used for each target language; these source languages are used for all experiments in the paper.
|82A||Order of subject and verb|
|83A||Order of object and verb|
|85A||Order of adposition and noun phrase|
|86A||Order of genitive and noun|
|87A||Order of adjective and noun|
|88A||Order of demonstrative and noun|
We use self-training [McClosky et al.2006] to further improve parsing performance. Specifically, we first train a delexicalized model on treebanks ; then use the resulting model to parse a dataset that includes target-language sentences which have POS tags but do not have dependency structures. We finally use the automatically parsed data as the treebank data and retrain the model; this last model is trained using all features (unlexicalized, clusters, and lexicalized). Self-training in this way gives an improvement in labeled accuracy from 63.18% to 63.91%.
|en||de, fr, pt, sv|
|de||en, fr, pt|
|es||fr, it, pt|
|fr||en, de, es, it, pt, sv|
|it||es, fr, pt|
|pt||en, de, es, fr, it, sv|
|sv||en, fr, pt|
2.4 Translation Dictionaries
Our only use of the translation data for is to construct a translation dictionary . Here and are two languages, is a word in language , and the output is a word in language corresponding to the most frequent translation of into this language.
We define the function as follows. We first run the GIZA++ alignment process [Och and Ney2000] on the data . We then keep intersected alignments between sentences in the two languages. Finally, for each word in , we define to be the target language word most frequently aligned to in the aligned data. If a word is never seen aligned to a target language word , we define .
3 Our Approach
We now describe an approach that gives significant improvements over the baseline. §3.1 describes a method for deriving cross-lingual clusters, allowing us to add cluster features to the model. §3.2 describes a method for adding lexical features to the model. §3.3 describes a method for integrating the approach with the density-driven approach of rasooli-collins:2015:EMNLP. Finally, §4 describes experiments. We show that each of the above steps leads to improvements in accuracy.
3.1 Learning Cross-Lingual Clusters
We now describe a method for learning cross-lingual clusters. This follows previous work on cross-lingual clustering algorithms [Täckström et al.2012]. A clustering is a function that maps each word in a vocabulary to a cluster , where is the number of clusters. A hierarchical clustering is a function that maps a word together with an integer to a cluster at level in the hierarchy. As one example, the Brown clustering algorithm [Brown et al.1992]
gives a hierarchical clustering. The levelallows cluster features at different levels of granularity.
A cross-lingual hierarchical clustering is a function where the clusters are shared across the languages of interest: that is, the word can be from any of the languages. Ideally, a cross-lingual clustering should put words across different languages which have a similar syntactic and/or semantic role in the same cluster. There is a clear motivation for cross-lingual clustering in the parsing context. We can use the cluster-based features on the source language treebanks , and these features will now generalize beyond these treebanks to the target language .
We learn a cross-lingual clustering by leveraging the monolingual data sets , together with the translation dictionaries learned from the translation data. Figure 1 shows the algorithm that learns a cross-lingual clustering. The algorithm first prepares a multilingual corpus, as follows: for each sentence in the monolingual data , for each word in
, with probabilitywe replace the word with its translation into some randomly chosen language. Once this data is created, we can easily obtain a cross-lingual clustering. Figure 1 shows the complete algorithm. The intuition behind this method is that by creating the cross-lingual data in this way, we bias the clustering algorithm towards putting words that are translations of each other in the same cluster.
3.2 Treebank Lexicalization
We now describe how to introduce lexical representations to the model. Our approach is simple: we take the treebank data for the
source languages, together with the translation lexicons. For any word in the source treebank data, we can look up its translation in the lexicon, and add this translated form to the underlying sentence. Features can now consider lexical identities derived in this way. In many cases the resulting translation will be the NULL word, leading to the absence of lexical features. However, the representations and still apply in this case, so the model is robust to some words having a NULL translation.
3.3 Integration with the Density-Driven Projection Method of rasooli-collins:2015:EMNLP
In this section we describe a method for integrating our approach with the cross-lingual transfer method of rasooli-collins:2015:EMNLP, which makes use of density-driven projections.
In annotation projection methods [Hwa et al.2005, McDonald et al.2011], it is assumed that we have translation data for a source and target language, and that we have a dependency parser in the source language . The translation data consists of pairs where is a source language sentence, and is a target language sentence. A method such as GIZA++ is used to derive an alignment between the words in and , for each sentence pair; the source language parser is used to parse . Each dependency in is then potentially transferred through the alignments to create a dependency in the target sentence . Once dependencies have been transferred in this way, a dependency parser can be trained on the dependencies in the target language.
The density-driven approach of rasooli-collins:2015:EMNLP makes use of various definitions of “density” of the projected dependencies. For example, is the set of projected structures where the projected dependencies form a full projective parse tree for the sentence; is the set of projected structures where at least 80% of the words in the projected structure are a modifier in some dependency. An iterative training process is used, where the parsing algorithm is first trained on the set of complete structures, and where progressively less dense structures are introduced in learning.
We integrate our approach with the density-driven approach of rasooli-collins:2015:EMNLP as follows: Consider the treebanks created using the lexicalization method of §3.2. We add all trees in these treebanks to the set of full trees used to initialize the method of rasooli-collins:2015:EMNLP. In addition we make use of the representations and , throughout the learning process.
This section first describes the experimental settings, then reports results.
4.1 Data and Tools
In a first set of experiments, we consider 7 European languages studied in several pieces of previous work [Ma and Xia2014, Zhang and Barzilay2015, Guo et al.2016, Ammar et al.2016, Lacroix et al.2016]. More specifically, we use the 7 European languages in the Google universal treebank (v.2; standard data) [McDonald et al.2013]. As in previous work, gold part-of-speech tags are used for evaluation. We use the concatenation of the treebank training sentences, Wikipedia data and the Bible monolingual sentences as our monolingual raw text. Table 3 shows statistics for the monolingual data. We use the Bible data from christodouloupoulos2014massively, which includes data for 100 languages, as the source of translations. We also conduct experiments with the Europarl data (both with the original size and a subset of it with the same size as the Bible text) to study the effects of translation data size and domain shift. The statistics for translation data is shown in Table 4.
In a second set of experiments, we ran experiments on 38 datasets (26 languages) in the more recent Universal Dependencies v1.3 corpus [Nivre et al.2016]. The full set of languages we use is listed in Table 9.444We excluded languages that are not completely present in the Bible corpus of christodouloupoulos2014massively (Ancient Greek, Basque, Catalan, Galician, Gothic, Irish, Kazakh, Latvian, Old Church Slavonic, and Tamil). We also excluded Arabic, Hebrew, Japanese and Chinese, as these languages have tokenization and/or morphological complexity that goes beyond the scope of this paper. Future work should consider these languages. We use the Bible as the translation data, and Wikipedia as the monolingual text. The standard training, development and test set splits are used in all experiments. The development sets are used for analysis, given in § 5 of this paper.
Brown Clustering Algorithm
We use the off-the-shelf Brown clustering tool555https://github.com/percyliang/brown-cluster [Liang2005] to train monolingual Brown clusters with 500 clusters. The monolingual Brown clusters are used as features over lexicalized values created in , and in self-training experiments. We train our cross-lingual clustering with the off-the-shelf-tool666https://github.com/karlstratos/singular from stratos-collins-hsu:2015:ACL-IJCNLP. We set the window size to 2 with cluster size of 500.777Usually the original Brown clusters are better features for parsing but their training procedure does not scale well to large datasets. Therefore we use the more efficient algorithm from stratos-collins-hsu:2015:ACL-IJCNLP on the larger cross-lingual datasets to obtain word clusters.
We use the k-beam arc-eager dependency parser of rasooli2015yara, which is similar to the model of zhang-nivre:2011:ACL-HLT2011. We modify the parser such that it can use both monolingual and cross-lingual word cluster features. The parser is trained using the the maximum violation update strategy [Huang et al.2012]
. We use three epochs of training for all experiments. We use the Dependable tool[Choi et al.2015] to calculate significance tests on several of the comparisons (details are given in the captions to tables 5, 6, and 9).
We use the intersected alignments from Giza++ [Och and Ney2000] on translation data. We exclude sentences in translation data with more than words.
4.2 Results on the Google Treebank
Table 5 shows the dependency parsing accuracy for the baseline delexicalized approach, and for models which add 1) cross-lingual clusters (§3.1); 2) lexical features (§3.2); 3) integration with the density-driven method of rasooli-collins:2015:EMNLP. It can be seen that each of these three steps gives significant improvements in performance. The final LAS/UAS of 73.9/80.3% is several percentage points higher than the baseline accuracy of 63.9/72.9%.
|L||Baseline||This paper using the Bible data|
|Density||This Paper||Density||This Paper||Density||This Paper|
Comparison to the Density-Driven Approach using Europarl Data
Table 6 shows accuracies for the density-driven approach of rasooli-collins:2015:EMNLP, first using Europarl data888rasooli-collins:2015:EMNLP do not report results on English. We use the same setting as in their paper to obtain the English results. and second using the Bible data alone (with no cross-lingual clusters or lexicalization). The Bible data is considerably smaller than Europarl (around 100 times smaller), and it can be seen that results using the Bible are several percentage points lower than the results for Europarl (75.7% UAS vs. 81.3% UAS). Integrating cluster and lexicalized features described in the current paper with the density-driven approach closes much of this gap in performance (80.3% UAS). Thus we have demonstrated that we can get close to the performance of the Europarl-based models using only the Bible as a source of translation data. Using our approach on the full Europarl data gives an average UAS of 82.9%, an improvement from the 81.3% UAS of rasooli-collins:2015:EMNLP.
Table 6 also shows results when we use a random subset of the Europarl data, in which the number of sentences (25,000) is chosen to give a very similar size to the Bible dataset. It can be seen that accuracies using the Bible data vs. Europarl-Sample are very similar (80.3% vs. 80.4% UAS), suggesting that the size of the translation corpus is much more important than the genre.
Comparison to Other Previous Work
Table 7 compares the accuracy of our method to the following related work: 1) ma-xia:2014:P14-1, who describe an annotation projection method based on entropy regularization; 2) lacroix-EtAl:2016:N16-1, who describe an annotation projection method based on training on partial trees with dynamic oracles; 3) yuanregina15, who describe a method that learns cross-lingual embeddings and bilingual dictionaries from Europarl data, and uses these features in a discriminative parsing model; 4) guo2016representation, who describe a method that learns cross-lingual embeddings from Europarl data and uses a shift-reduce neural parser with these representations; 5) ammar2016one, who use the same embeddings as guo2016representation, within an LSTM-based parser; and 6) rasooli-collins:2015:EMNLP who use the density-driven approach on the Europarl data. Our method gives significant improvements over the first three models, in spite of using the Bible translation data rather than Europarl. When using the Europarl data, our method improves the state-of-the-art model of rasooli-collins:2015:EMNLP.
|Lang.||RC15||This Paper (§3.3)|
Performance with Automatic POS Tags
For completeness, Table 8 gives results for our method with automatic part-of-speech tags. The tags are obtained using the model of collins:2002:EMNLP02999https://github.com/rasoolims/SemiSupervisedPosTagger trained on the training part of the treebank dataset. Future work should study approaches that transfer POS tags in addition to dependencies.
4.3 Results on the Universal Dependencies v1.3
Table 9 gives results on 38 datasets (26 languages) from the newly released universal dependencies corpus [Nivre et al.2016]. Given the number of treebanks and to speed up training, we pick source languages that have at least 5 out of 6 common WALS properties with each target language. Our experiments are carried out using the Bible as our translation data. As shown in Table 9, our method consistently outperforms the density-driven method of rasooli-collins:2015:EMNLP and for many languages the accuracy of our method gets close to the accuracy of the supervised parser. In all the languages, our method is significantly better than the density-driven method using the McNemar’s test with .
Accuracy on some languages (e.g., Persian (fa) and Turkish (tr)) is low, suggesting that future work should consider more powerful techniques for these languages. There are two important facts to note. First, the number of fully projected trees in some languages is so low such that the density-driven approach cannot start with a good initialization to fill in partial dependencies. For example Turkish has only one full tree with only six words, Persian with 25 trees, and Dutch with 28 trees. Second, we observe very low accuracies in supervised parsing for some languages in which the number of training sentences is very low (for example, Latin has only 1326 projective trees in the training data).
Precision, recall and f-score of different dependency relations on the English development data of the Google universal treebank. The major columns show the dependency labels (“dep.”), frequency (“freq.”), the baseline delexicalized model (“delex”), and our method using the Bible and Europarl (“EU”) as translation data. The rows are sorted by frequency.
We conclude with some analysis of the accuracy of the method on different dependency types, across the different languages. Table 10
shows precision and recall on different dependency types in English (using the Google treebank). The improvements in accuracy when moving from the delexicalized model to the Bible or Europarl model apply quite uniformly across all dependency types, with all dependency labels showing an improvement.
Table 11 shows the dependency accuracy sorted by part-of-speech tag of the modifier in the dependency. We break the results into three groups: G1 languages, where UAS is at least 80% overall; G2 languages, where UAS is between 70% and 80%; and G3 languages, where UAS is less than 70%. There are some quite significant differences in accuracy depending on the POS of the modifier word: for example in the G1 languages ADP, DET, ADJ, PRON and AUX all have over 85% accuracy; in contrast NOUN, VERB, PROPN, ADV all have accuracy that is less than 80%. A very similar pattern is seen for the G2 languages, with ADP, DET, ADJ, and AUX again having greater than 85% accuracy, but NOUN, VERB, PROPN and ADV having lower accuracies. These results suggest that difficulty varies quite signficantly depending on the modifier POS, and different languages show the same patterns of difficulty with respect to modifier POS.
Table 12 shows accuracy sorted by the POS tag of the head word of the dependency. By far the most frequent head POS tags are NOUN, VERB, and PROPN (accounting for 85% of all dependencies). The table also shows that for all language groups G1, G2, and G3, the f1 scores for NOUN, VERB and PROPN are generally higher than the f1 scores for other head POS tags.
Finally, Table 13 shows precision and recall for different dependency labels for the G1, G2 and G3 languages. We again see quite large differences in accuracy between different dependency labels. For example in the G1 languages dependencies with the nmod label, the most frequent label, have 75.2% f1 score; in contrast the second most frequent label, case, has 93.7% f1 score. Other frequent labels with low accuracy in the G1 languages are advmod, conj, and cc.
6 Related Work
There has recently been a great deal of work on syntactic transfer. A number of methods [Zeman and Resnik2008, McDonald et al.2011, Cohen et al.2011, Naseem et al.2012, Täckström et al.2013, Rosa and Zabokrtsky2015] directly learn delexicalized models that can be trained on universal treebank data from one or more source languages, then applied to the target language. More recent work has introduced cross-lingual representations—for example cross-lingual word-embeddings—that can be used to improve performance [Zhang and Barzilay2015, Guo et al.2015, Duong et al.2015a, Duong et al.2015b, Guo et al.2016, Ammar et al.2016]. These cross-lingual representations are usually learned from parallel translation data. We show results for the methods of [Zhang and Barzilay2015, Guo et al.2016, Ammar et al.2016] in Table 7 of this paper.
The annotation projection approach, where dependencies from one language are transferred through translation alignments to another language, has been considered by several authors [Hwa et al.2005, Ganchev et al.2009, McDonald et al.2011, Ma and Xia2014, Rasooli and Collins2015, Lacroix et al.2016, Agić et al.2016, Schlichtkrull and Søgaard2017].
Other recent work [Tiedemann et al.2014, Tiedemann2015, Tiedemann and Agić2016] has considered treebank translation, where a statistical machine translation system (e.g., MOSES [Koehn et al.2007]) is used to translate a source language treebank into the target language, complete with reordering of the input sentence. The lexicalization approach described in this paper is a simple form of treebank translation, where we use a word-to-word translation model. In spite of its simplicity, it is an effective approach.
A number of authors have considered incorporating universal syntactic properties such as dependency order by selectively learning syntactic attributes from similar source languages [Naseem et al.2012, Täckström et al.2013, Zhang and Barzilay2015, Ammar et al.2016]. Selective sharing of syntactic properties is complementary to our work; we used a very limited form of selective sharing, through the WALS properties, in our baseline approach. More recently, wang2016galactic have developed a synthetic treebank as a universal treebank to help learn parsers for new languages. alonso2017parsing try a very different approach in cross-lingual transfer by using a ranking approach.
A number of authors [Täckström et al.2012, Guo et al.2015, Guo et al.2016] have introduced methods that learn cross-lingual representations that are then used in syntactic transfer. Most of these approaches introduce constraints to a clustering or embedding algorithm that encourage words that are translations of each other to have similar representations. Our method of deriving a cross-lingual corpus (see Figure 1) is closely related to [Duong et al.2015a, Gouws and Søgaard2015, Wick et al.2015].
Our work has made use of dictionaries that are automatically extracted from bilingual corpora. An alternative approach would be to use hand-crafted translation lexicons, for example PanLex [Baldwin et al.2010], which covers 1253 language varieties, or Wiktionary (e.g., see [Durrett et al.2012] for an approach that uses Wiktionary for cross-lingual transfer). These resources are a potentially very rich source of information; future work should investigate whether they can give improvements in performance.
We have described a method for cross-lingual syntactic transfer that is effective in the scenario where a large amount of translation data is not available. We have introduced a simple, direct method for deriving cross-lingual clusters, and for transferring lexical information across treebanks for different languages. Experiments with the method show that the method gives improved performance over previous work that makes use of Europarl, a much larger translation corpus.
Appendix A Parsing Features
We used all features in [Zhang and Nivre2011, Table 1 and 2], which describes features based on the word and part-of-speech at various positions on the stack and buffer of the transition system. In addition, we expand the [Table 1]zhang-nivre:2011:ACL-HLT2011 features to include clusters, as follows: whenever a feature tests the part-of-speech for a word in position 0 of the stack or buffer, we introduce features that replace the part-of-speech with the Brown clustering bit-string of length 4 and 6. Whenever a feature tests for the word identity at position 0 of the stack or buffer, we introduce a cluster feature that replaces the word with the full cluster feature. We take the cross product of all features corresponding to the choice of 4 or 6 length bit string for part-of-speech features.
- [Agić et al.2016] Željko Agić, Anders Johannsen, Barbara Plank, Héctor Alonso Martínez, Natalie Schluter, and Anders Søgaard. 2016. Multilingual projection for parsing truly low-resource languages. Transactions of the Association for Computational Linguistics, 4:301–312.
- [Alonso et al.2017] Héctor Martínez Alonso, Željko Agić, Barbara Plank, and Anders Søgaard. 2017. Parsing universal dependencies without training. arXiv preprint arXiv:1701.03163.
- [Ammar et al.2016] Waleed Ammar, George Mulcaire, Miguel Ballesteros, Chris Dyer, and Noah A Smith. 2016. One parser, many languages. arXiv preprint arXiv:1602.01595.
- [Baldwin et al.2010] Timothy Baldwin, Jonathan Pool, and Susan M Colowick. 2010. Panlex and lextract: Translating all words of all languages of the world. In Proceedings of the 23rd International Conference on Computational Linguistics: Demonstrations, pages 37–40. Association for Computational Linguistics.
[Brown et al.1992]
Peter F Brown, Peter V Desouza, Robert L Mercer, Vincent J Della Pietra, and
Jenifer C Lai.
Class-based n-gram models of natural language.Computational linguistics, 18(4):467–479.
[Choi et al.2015]
Jinho D Choi, Joel Tetreault, and Amanda Stent.
It depends: Dependency parser comparison using a web-based evaluation
Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing of the Asian Federation of Natural Language Processing, ACL, pages 26–31.
- [Christodouloupoulos and Steedman2014] Christos Christodouloupoulos and Mark Steedman. 2014. A massively parallel corpus: the bible in 100 languages. Language Resources and Evaluation, pages 1–21.
- [Cohen et al.2011] Shay B. Cohen, Dipanjan Das, and Noah A. Smith. 2011. Unsupervised structure prediction with non-parallel multilingual guidance. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 50–61, Edinburgh, Scotland, UK., July. Association for Computational Linguistics.
Discriminative training methods for hidden markov models: Theory and experiments with perceptron algorithms.In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing, pages 1–8. Association for Computational Linguistics, July.
- [Dryer and Haspelmath2013] Matthew S. Dryer and Martin Haspelmath, editors. 2013. WALS Online. Max Planck Institute for Evolutionary Anthropology, Leipzig.
- [Duong et al.2015a] Long Duong, Trevor Cohn, Steven Bird, and Paul Cook. 2015a. Cross-lingual transfer for unsupervised dependency parsing without parallel data. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning, pages 113–122, Beijing, China, July. Association for Computational Linguistics.
[Duong et al.2015b]
Long Duong, Trevor Cohn, Steven Bird, and Paul Cook.
A neural network model for low-resource universal dependency parsing.In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 339–348, Lisbon, Portugal, September. Association for Computational Linguistics.
- [Durrett et al.2012] Greg Durrett, Adam Pauls, and Dan Klein. 2012. Syntactic transfer using a bilingual lexicon. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 1–11, Jeju Island, Korea, July. Association for Computational Linguistics.
- [Ganchev et al.2009] Kuzman Ganchev, Jennifer Gillenwater, and Ben Taskar. 2009. Dependency grammar induction via bitext projection constraints. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP, pages 369–377, Suntec, Singapore, August. Association for Computational Linguistics.
- [Gouws and Søgaard2015] Stephan Gouws and Anders Søgaard. 2015. Simple task-specific bilingual word embeddings. In Proceedings of the 2015 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1386–1390, Denver, Colorado, May–June. Association for Computational Linguistics.
[Guo et al.2015]
Jiang Guo, Wanxiang Che, David Yarowsky, Haifeng Wang, and Ting Liu.
Cross-lingual dependency parsing based on distributed representations.In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1234–1244, Beijing, China, July. Association for Computational Linguistics.
[Guo et al.2016]
Jiang Guo, Wanxiang Che, David Yarowsky, Haifeng Wang, and Ting Liu.
A representation learning framework for multi-source transfer
The Thirtieth AAAI Conference on Artificial Intelligence (AAAI-16), Phoenix, Arizona, USA.
- [Huang et al.2012] Liang Huang, Suphan Fayong, and Yang Guo. 2012. Structured perceptron with inexact search. In Proceedings of the 2012 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 142–151, Montréal, Canada, June. Association for Computational Linguistics.
- [Hwa et al.2005] Rebecca Hwa, Philip Resnik, Amy Weinberg, Clara Cabezas, and Okan Kolak. 2005. Bootstrapping parsers via syntactic projection across parallel texts. Natural language engineering, 11(03):311–325.
- [Koehn et al.2007] Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, et al. 2007. Moses: Open source toolkit for statistical machine translation. In Proceedings of the 45th annual meeting of the ACL on interactive poster and demonstration sessions, pages 177–180. Association for Computational Linguistics.
- [Koehn2005] Philipp Koehn. 2005. Europarl: A parallel corpus for statistical machine translation. In MT summit, volume 5, pages 79–86.
- [Lacroix et al.2016] Ophélie Lacroix, Lauriane Aufrant, Guillaume Wisniewski, and François Yvon. 2016. Frustratingly easy cross-lingual transfer for transition-based dependency parsing. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1058–1063, San Diego, California, June. Association for Computational Linguistics.
- [Liang2005] Percy Liang. 2005. Semi-supervised learning for natural language. Ph.D. thesis, Massachusetts Institute of Technology.
- [Ma and Xia2014] Xuezhe Ma and Fei Xia. 2014. Unsupervised dependency parsing with transferring distribution via parallel guidance and entropy regularization. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1337–1348, Baltimore, Maryland, June. Association for Computational Linguistics.
- [McClosky et al.2006] David McClosky, Eugene Charniak, and Mark Johnson. 2006. Effective self-training for parsing. In Proceedings of the Main Conference on Human Language Technology Conference of the North American Chapter of the Association of Computational Linguistics, HLT-NAACL ’06, pages 152–159, Stroudsburg, PA, USA. Association for Computational Linguistics.
- [McDonald et al.2011] Ryan McDonald, Slav Petrov, and Keith Hall. 2011. Multi-source transfer of delexicalized dependency parsers. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 62–72, Edinburgh, Scotland, UK., July. Association for Computational Linguistics.
- [McDonald et al.2013] Ryan McDonald, Joakim Nivre, Yvonne Quirmbach-Brundage, Yoav Goldberg, Dipanjan Das, Kuzman Ganchev, Keith Hall, Slav Petrov, Hao Zhang, Oscar Täckström, Claudia Bedini, Núria Bertomeu Castelló, and Jungmee Lee. 2013. Universal dependency annotation for multilingual parsing. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 92–97, Sofia, Bulgaria, August. Association for Computational Linguistics.
- [Naseem et al.2012] Tahira Naseem, Regina Barzilay, and Amir Globerson. 2012. Selective sharing for multilingual dependency parsing. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics: Long Papers-Volume 1, pages 629–637. Association for Computational Linguistics.
- [Nivre et al.2016] Joakim Nivre, Željko Agić, Lars Ahrenberg, Maria Jesus Aranzabe, Masayuki Asahara, Aitziber Atutxa, Miguel Ballesteros, John Bauer, Kepa Bengoetxea, Yevgeni Berzak, Riyaz Ahmad Bhat, Cristina Bosco, Gosse Bouma, Sam Bowman, Gülşen Cebiroğlu Eryiğit, Giuseppe G. A. Celano, Çağrı Çöltekin, Miriam Connor, Marie-Catherine de Marneffe, Arantza Diaz de Ilarraza, Kaja Dobrovoljc, Timothy Dozat, Kira Droganova, Tomaž Erjavec, Richárd Farkas, Jennifer Foster, Daniel Galbraith, Sebastian Garza, Filip Ginter, Iakes Goenaga, Koldo Gojenola, Memduh Gokirmak, Yoav Goldberg, Xavier Gómez Guinovart, Berta Gonzáles Saavedra, Normunds Grūzītis, Bruno Guillaume, Jan Hajič, Dag Haug, Barbora Hladká, Radu Ion, Elena Irimia, Anders Johannsen, Hüner Kaşıkara, Hiroshi Kanayama, Jenna Kanerva, Boris Katz, Jessica Kenney, Simon Krek, Veronika Laippala, Lucia Lam, Alessandro Lenci, Nikola Ljubešić, Olga Lyashevskaya, Teresa Lynn, Aibek Makazhanov, Christopher Manning, Cătălina Mărănduc, David Mareček, Héctor Martínez Alonso, Jan Mašek, Yuji Matsumoto, Ryan McDonald, Anna Missilä, Verginica Mititelu, Yusuke Miyao, Simonetta Montemagni, Keiko Sophie Mori, Shunsuke Mori, Kadri Muischnek, Nina Mustafina, Kaili Müürisep, Vitaly Nikolaev, Hanna Nurmi, Petya Osenova, Lilja Øvrelid, Elena Pascual, Marco Passarotti, Cenel-Augusto Perez, Slav Petrov, Jussi Piitulainen, Barbara Plank, Martin Popel, Lauma Pretkalniņa, Prokopis Prokopidis, Tiina Puolakainen, Sampo Pyysalo, Loganathan Ramasamy, Laura Rituma, Rudolf Rosa, Shadi Saleh, Baiba Saulīte, Sebastian Schuster, Wolfgang Seeker, Mojgan Seraji, Lena Shakurova, Mo Shen, Natalia Silveira, Maria Simi, Radu Simionescu, Katalin Simkó, Kiril Simov, Aaron Smith, Carolyn Spadine, Alane Suhr, Umut Sulubacak, Zsolt Szántó, Takaaki Tanaka, Reut Tsarfaty, Francis Tyers, Sumire Uematsu, Larraitz Uria, Gertjan van Noord, Viktor Varga, Veronika Vincze, Jing Xian Wang, Jonathan North Washington, Zdeněk Žabokrtský, Daniel Zeman, and Hanzhi Zhu. 2016. Universal dependencies 1.3. LINDAT/CLARIN digital library at Institute of Formal and Applied Linguistics, Charles University in Prague.
- [Och and Ney2000] Franz Josef Och and Hermann Ney. 2000. Giza++: Training of statistical translation models.
- [Rasooli and Collins2015] Mohammad Sadegh Rasooli and Michael Collins. 2015. Density-driven cross-lingual transfer of dependency parsers. In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 328–338, Lisbon, Portugal, September. Association for Computational Linguistics.
- [Rasooli and Tetreault2015] Mohammad Sadegh Rasooli and Joel Tetreault. 2015. Yara parser: A fast and accurate dependency parser. arXiv preprint arXiv:1503.06733.
- [Rosa and Zabokrtsky2015] Rudolf Rosa and Zdenek Zabokrtsky. 2015. Klcpos3 - a language similarity measure for delexicalized parser transfer. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 243–249, Beijing, China, July. Association for Computational Linguistics.
- [Schlichtkrull and Søgaard2017] Michael Sejr Schlichtkrull and Anders Søgaard. 2017. Cross-lingual dependency parsing with late decoding for truly low-resource languages. arXiv preprint arXiv:1701.01623.
- [Stratos et al.2015] Karl Stratos, Michael Collins, and Daniel Hsu. 2015. Model-based word embeddings from decompositions of count matrices. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 1282–1291, Beijing, China, July. Association for Computational Linguistics.
- [Täckström et al.2012] Oscar Täckström, Ryan McDonald, and Jakob Uszkoreit. 2012. Cross-lingual word clusters for direct transfer of linguistic structure. In Proceedings of the 2012 conference of the North American chapter of the association for computational linguistics: Human language technologies, pages 477–487. Association for Computational Linguistics.
- [Täckström et al.2013] Oscar Täckström, Ryan McDonald, and Joakim Nivre. 2013. Target language adaptation of discriminative transfer parsers. Transactions for ACL.
- [Tiedemann and Agić2016] Jörg Tiedemann and Željko Agić. 2016. Synthetic treebanking for cross-lingual dependency parsing. Journal of Artificial Intelligence Research, 55:209–248.
- [Tiedemann et al.2014] Jörg Tiedemann, Željko Agić, and Joakim Nivre. 2014. Treebank translation for cross-lingual parser induction. In Proceedings of the Eighteenth Conference on Computational Natural Language Learning, pages 130–140, Ann Arbor, Michigan, June. Association for Computational Linguistics.
- [Tiedemann2015] Jörg Tiedemann. 2015. Improving the cross-lingual projection of syntactic dependencies. In Nordic Conference of Computational Linguistics NODALIDA 2015, pages 191–199.
- [Wang and Eisner2016] Dingquan Wang and Jason Eisner. 2016. The galactic dependencies treebanks: Getting more data by synthesizing new languages. Transactions of the Association for Computational Linguistics, 4:491–505.
- [Wick et al.2015] Michael Wick, Pallika Kanani, and Adam Pocock. 2015. Minimally-constrained multilingual embeddings via artificial code-switching. In Workshop on Transfer and Multi-Task Learning: Trends and New Perspectives, Montreal, Canada, December.
- [Xiao and Guo2015] Min Xiao and Yuhong Guo. 2015. Annotation projection-based representation learning for cross-lingual dependency parsing. In Proceedings of the Nineteenth Conference on Computational Natural Language Learning, pages 73–82, Beijing, China, July. Association for Computational Linguistics.
- [Zeman and Resnik2008] Daniel Zeman and Philip Resnik. 2008. Cross-language parser adaptation between related languages. In Proceedings of the IJCNLP-08 Workshop on NLP for Less Privileged Languages, pages 35–42.
[Zhang and Barzilay2015]
Yuan Zhang and Regina Barzilay.
Hierarchical low-rank tensors for multilingual transfer parsing.In Conference on Empirical Methods in Natural Language Processing (EMNLP), Lisbon, Portugal, September.
- [Zhang and Nivre2011] Yue Zhang and Joakim Nivre. 2011. Transition-based dependency parsing with rich non-local features. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 188–193, Portland, Oregon, USA, June. Association for Computational Linguistics.