Elamur Research⌕
☰
home / papers / 2013 / A Hybrid Approach for Anaphora Resolution in Hindi

A Hybrid Approach for Anaphora Resolution in Hindi

Summary

The paper tackles automatic resolution of pronouns that refer to concrete entities in Hindi. Most earlier work relied on phrase-structure syntax. The authors instead use dependency analyses, which suit Hindi's relatively free word order. Their data comes from the Hindi/Urdu Dependency Treebank, annotated in the Computational Paninian Grammar (CPG) framework with karaka relations, plus animacy labels, Named Entity tags and anaphora links.

The system works in two stages. First, category-specific rules handle reflexive, relative, locative and personal pronouns by walking the dependency tree. For example, a reflexive is linked to the k1 dependent of its clause's verb, and a relative pronoun to the noun its nmod-relc clause modifies. Pronouns the rules cannot settle go to a decision-tree classifier that ranks candidate antecedents pairwise, using distance, agreement, animacy and NE features.

The rules alone reach 0.60 accuracy and are weak mainly on third-person pronouns. The full hybrid reaches 0.70. The work shows that dependency structure plus light semantic cues gives a practical route to anaphora resolution for Indian languages.

Key points

  • Data: 325 treebank documents with 4970 pronouns, of which 3233 are entity pronouns, split roughly 2:1 into 2162 training and 1071 test pronouns.
  • Rule-based accuracy by pronoun type:
    • first/second person: 0.93
    • relative: 0.85
    • reflexive: 0.82
    • locative: 0.77
    • third person: 0.48
    • overall: 0.60 (653/1071)
  • Third-person pronouns are handled with a salience ordering of dependency relations (k1 > k2 > r6 > k4 > k3 > others), with candidates filtered by number and animacy.
  • The classifier follows the Soon et al. (2001) pairwise scheme, considering candidates from up to 3 previous sentences. Adding distance, animacy and agreement features raises accuracy to 0.70 (753/1071).
  • The authors give a tentative comparison, since the earlier systems' data was unavailable:
    • Dakwale and Sharma (2011): 0.52 on 258 pronouns
    • Uppalapu and Sharma (2009): 0.86 and 0.64 on 142 and 100 pronouns

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Topics: Coreference resolution

Cited by

7 papers of our library cite it (of the 140673 whose bibliographies we read); 2 of them have a page here, 5 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{dakwale2013hybrid,
  title = {{A Hybrid Approach for Anaphora Resolution in Hindi}},
  author = {Dakwale, Praveen and Mujadia, Vandan and Sharma, Dipti M},
  booktitle = {Proceedings of the Sixth International Joint Conference on Natural Language Processing},
  year = {2013},
  url = {https://aclanthology.org/I13-1130/}
}

References

14 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rafiya Begum, Samar Husain, Arun Dhwaj, Dipti Misra Sharma, Lakshmi Bai, and Rajeev Sangal. 2008. Dependency annotation scheme for indian languages. In Proceedings of IJCNLP. find
  2. Akshar Bharati, Vineet Chaitanya, Rajeev Sangal, and KV Ramakrishnamacharyulu. 1995. Natural language processing: a Paninian perspective. Prentice-Hall of India. find
  3. Rajesh Bhatt, Bhuvana Narasimhan, Martha Palmer, Owen Rambow, Dipti Misra Sharma, and Fei Xia. 2009. A multi-representational and multi-layered treebank for hindi/urdu. In Proceedings of the LAWIII. ACL. find
  4. Anders Björkelund and Jonas Kuhn. 2012. Phrase structures and dependencies for end-to-end coreference resolution. In Proceedings of COLING 2012. find
  5. Susan E Brennan, Marilyn W Friedman, and Carl J Pollard. 1987. A centering approach to pronouns. In Proceedings of 25th ACL. find
  6. Praveen Dakwale and Himanshu Sharma. 2011. Anaphora resolution in indian languages using hybrid approaches. In NLP tool contest, ICON 2011. find
  7. Praveen Dakwale, Himanshu Sharma, and Dipti M Sharma. 2012. Anaphora annotation in hindi dependency treebank. In Proceedings of PACLIC-26. find
  8. Jerry Hobbs. 1986. Resolving pronoun references. In Readings in natural language processing. Morgan Kaufmann Publishers Inc. find
  9. Itisree Jena, Riyaz Ahmad Bhat, and Sambhav Jain. 2013. Animacy annotation in hindi treebank. In Proceedings of the LAWVII. ACL. find
  10. Igor Aleksandrovič Melčuk. 1988. Dependency syntax: theory and practice. State University of New York Press. find
  11. Rashmi Prasad and Michael Strube. 2000. Discourse salience and pronoun resolution in hindi. U. Penn Working Papers in Linguistics. find
  12. Wee Meng Soon, Hwee Tou Ng, and Daniel Chung Yong Lim. 2001. A machine learning approach to coreference resolution of noun phrases. Computational linguistics, 27. find our card
  13. Bhargav Uppalapu and Dipti Misra Sharma. Pronoun resolution for hindi. In DAARC-7. 2009. find
  14. Ian H Witten and Eibe Frank. 2005. Data Mining: Practical machine learning tools and techniques. Morgan Kaufmann. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/dakwale-2013-hybrid-approach-anaphora-resolution-hindi/