Cross-lingual Structure Transfer for Relation and Event Extraction

Cross-lingual Structure Transfer for Relation and Event Extraction. Subburathinam, A., Lu, D., Ji, H., May, J., Chang, S., Sil, A., & Voss, C. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), pages 313–325, Hong Kong, China, November, 2019. Association for Computational Linguistics.

Paper doi abstract bibtex

The identification of complex semantic structures such as events and entity relations, already a challenging Information Extraction task, is doubly difficult from sources written in under-resourced and under-annotated languages. We investigate the suitability of cross-lingual structure transfer techniques for these tasks. We exploit relation- and event-relevant language-universal features, leveraging both symbolic (including part-of-speech and dependency path) and distributional (including type representation and contextualized representation) information. By representing all entity mentions, event triggers, and contexts into this complex and structured multilingual common space, using graph convolutional networks, we can train a relation or event extractor from source language annotations and apply it to the target language. Extensive experiments on cross-lingual relation and event transfer among English, Chinese, and Arabic demonstrate that our approach achieves performance comparable to state-of-the-art supervised models trained on up to 3,000 manually annotated mentions: up to 62.6% F-score for Relation Extraction, and 63.1% F-score for Event Argument Role Labeling. The event argument role labeling model transferred from English to Chinese achieves similar performance as the model trained from Chinese. We thus find that language-universal symbolic and distributional representations are complementary for cross-lingual structure transfer.

@inproceedings{subburathinam-etal-2019-cross,
    title = "Cross-lingual Structure Transfer for Relation and Event Extraction",
    author = "Subburathinam, Ananya  and
      Lu, Di  and
      Ji, Heng  and
      May, Jonathan  and
      Chang, Shih-Fu  and
      Sil, Avirup  and
      Voss, Clare",
    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)",
    month = nov,
    year = "2019",
    address = "Hong Kong, China",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/D19-1030",
    doi = "10.18653/v1/D19-1030",
    pages = "313--325",
    abstract = "The identification of complex semantic structures such as events and entity relations, already a challenging Information Extraction task, is doubly difficult from sources written in under-resourced and under-annotated languages. We investigate the suitability of cross-lingual structure transfer techniques for these tasks. We exploit relation- and event-relevant language-universal features, leveraging both symbolic (including part-of-speech and dependency path) and distributional (including type representation and contextualized representation) information. By representing all entity mentions, event triggers, and contexts into this complex and structured multilingual common space, using graph convolutional networks, we can train a relation or event extractor from source language annotations and apply it to the target language. Extensive experiments on cross-lingual relation and event transfer among English, Chinese, and Arabic demonstrate that our approach achieves performance comparable to state-of-the-art supervised models trained on up to 3,000 manually annotated mentions: up to 62.6{\%} F-score for Relation Extraction, and 63.1{\%} F-score for Event Argument Role Labeling. The event argument role labeling model transferred from English to Chinese achieves similar performance as the model trained from Chinese. We thus find that language-universal symbolic and distributional representations are complementary for cross-lingual structure transfer.",
}

Downloads: 0

{"_id":"22dQ9AKWDMaaHZGqx","bibbaseid":"subburathinam-lu-ji-may-chang-sil-voss-crosslingualstructuretransferforrelationandeventextraction-2019","author_short":["Subburathinam, A.","Lu, D.","Ji, H.","May, J.","Chang, S.","Sil, A.","Voss, C."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"Cross-lingual Structure Transfer for Relation and Event Extraction","author":[{"propositions":[],"lastnames":["Subburathinam"],"firstnames":["Ananya"],"suffixes":[]},{"propositions":[],"lastnames":["Lu"],"firstnames":["Di"],"suffixes":[]},{"propositions":[],"lastnames":["Ji"],"firstnames":["Heng"],"suffixes":[]},{"propositions":[],"lastnames":["May"],"firstnames":["Jonathan"],"suffixes":[]},{"propositions":[],"lastnames":["Chang"],"firstnames":["Shih-Fu"],"suffixes":[]},{"propositions":[],"lastnames":["Sil"],"firstnames":["Avirup"],"suffixes":[]},{"propositions":[],"lastnames":["Voss"],"firstnames":["Clare"],"suffixes":[]}],"booktitle":"Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)","month":"November","year":"2019","address":"Hong Kong, China","publisher":"Association for Computational Linguistics","url":"https://www.aclweb.org/anthology/D19-1030","doi":"10.18653/v1/D19-1030","pages":"313–325","abstract":"The identification of complex semantic structures such as events and entity relations, already a challenging Information Extraction task, is doubly difficult from sources written in under-resourced and under-annotated languages. We investigate the suitability of cross-lingual structure transfer techniques for these tasks. We exploit relation- and event-relevant language-universal features, leveraging both symbolic (including part-of-speech and dependency path) and distributional (including type representation and contextualized representation) information. By representing all entity mentions, event triggers, and contexts into this complex and structured multilingual common space, using graph convolutional networks, we can train a relation or event extractor from source language annotations and apply it to the target language. Extensive experiments on cross-lingual relation and event transfer among English, Chinese, and Arabic demonstrate that our approach achieves performance comparable to state-of-the-art supervised models trained on up to 3,000 manually annotated mentions: up to 62.6% F-score for Relation Extraction, and 63.1% F-score for Event Argument Role Labeling. The event argument role labeling model transferred from English to Chinese achieves similar performance as the model trained from Chinese. We thus find that language-universal symbolic and distributional representations are complementary for cross-lingual structure transfer.","bibtex":"@inproceedings{subburathinam-etal-2019-cross,\n title = \"Cross-lingual Structure Transfer for Relation and Event Extraction\",\n author = \"Subburathinam, Ananya and\n Lu, Di and\n Ji, Heng and\n May, Jonathan and\n Chang, Shih-Fu and\n Sil, Avirup and\n Voss, Clare\",\n booktitle = \"Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP)\",\n month = nov,\n year = \"2019\",\n address = \"Hong Kong, China\",\n publisher = \"Association for Computational Linguistics\",\n url = \"https://www.aclweb.org/anthology/D19-1030\",\n doi = \"10.18653/v1/D19-1030\",\n pages = \"313--325\",\n abstract = \"The identification of complex semantic structures such as events and entity relations, already a challenging Information Extraction task, is doubly difficult from sources written in under-resourced and under-annotated languages. We investigate the suitability of cross-lingual structure transfer techniques for these tasks. We exploit relation- and event-relevant language-universal features, leveraging both symbolic (including part-of-speech and dependency path) and distributional (including type representation and contextualized representation) information. By representing all entity mentions, event triggers, and contexts into this complex and structured multilingual common space, using graph convolutional networks, we can train a relation or event extractor from source language annotations and apply it to the target language. Extensive experiments on cross-lingual relation and event transfer among English, Chinese, and Arabic demonstrate that our approach achieves performance comparable to state-of-the-art supervised models trained on up to 3,000 manually annotated mentions: up to 62.6{\\%} F-score for Relation Extraction, and 63.1{\\%} F-score for Event Argument Role Labeling. The event argument role labeling model transferred from English to Chinese achieves similar performance as the model trained from Chinese. We thus find that language-universal symbolic and distributional representations are complementary for cross-lingual structure transfer.\",\n}\n\n","author_short":["Subburathinam, A.","Lu, D.","Ji, H.","May, J.","Chang, S.","Sil, A.","Voss, C."],"key":"subburathinam-etal-2019-cross","id":"subburathinam-etal-2019-cross","bibbaseid":"subburathinam-lu-ji-may-chang-sil-voss-crosslingualstructuretransferforrelationandeventextraction-2019","role":"author","urls":{"Paper":"https://www.aclweb.org/anthology/D19-1030"},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://jonmay.github.io/webpage/cutelabname/cutelabname.bib","dataSources":["ZdhKtP2cSp3Aki2ge","X5WBAKQabka5TW5z7","BnZgtH7HDESgbxKxt","hbZSwot2msWk92m5B","fcWjcoAgajPvXWcp7","GvHfaAWP6AfN6oLQE","j3Qzx9HAAC6WtJDHS","5eM3sAccSEpjSDHHQ"],"keywords":[],"search_terms":["cross","lingual","structure","transfer","relation","event","extraction","subburathinam","lu","ji","may","chang","sil","voss"],"title":"Cross-lingual Structure Transfer for Relation and Event Extraction","year":2019}