Elamur Research⌕
☰
home / papers / 2009 / The Hindi Discourse Relation Bank

The Hindi Discourse Relation Bank

Summary

The paper presents the Hindi Discourse Relation Bank (HDRB), an ongoing effort to build a Hindi corpus annotated with discourse relations, modelled on the lexically grounded Penn Discourse Treebank (PDTB). HDRB targets 200K words of Amar Ujala news text, a subset of a 400K-word corpus that is separately receiving syntactic dependency annotation.

The authors adapt PDTB to Hindi in several ways:

  • Connectives: they add new connective classes.
  • Argument labels: Arg1 and Arg2 are assigned by the semantic role each argument plays in a given sense, not by syntax.
  • Implicit relations: these are also annotated across paragraph boundaries.
  • Sense hierarchy: they revise it, including a uniform treatment of pragmatic relations.

A pilot annotation of 35 texts gave 602 relation tokens. In this sample, Hindi's rich morphology did not reduce its reliance on explicit connectives. Hindi did use more alternative lexicalizations (AltLex) than English. The share of each sense class realized explicitly versus implicitly resembled PDTB.

The resource extends discourse annotation to a major Indian language and supports cross-linguistic study of discourse relations.

Key points

  • Corpus: HDRB covers 200K words of Hindi newspaper text (politics, sports, films, etc.). It is drawn from a 400K-word corpus with dependency annotation (Begum et al., 2008).
  • New connective classes: Beyond PDTB's subordinating conjunctions, coordinating conjunctions and adverbials, HDRB adds:
    • sentential relatives (e.g., जिससे "so that");
    • subordinators, whose annotation is postponed to a later phase;
    • particles (e.g., भी).
  • Sense scheme changes:
    • Subtype labels that only encoded argument order were removed.
    • A new "Goal" type was added under Contingency.
    • Pragmatic senses are split into epistemic, speech-act and propositional subtypes.
  • Pilot relation counts (35 texts, 602 tokens):
    • Explicit: 189 tokens (31.4%), 49 unique connectives, about half the PDTB count from 1M words.
    • Implicit: 185 tokens (30.7%).
    • AltLex: 37 tokens (6.14%), versus 1.5% in PDTB.
    • EntRel: 140 tokens (23.25%).
    • NoRel: 51 tokens (8.5%).
  • Explicit realization by sense class: Comparison 76.5%, Temporal 65.2%, Contingency 58.2%, Expansion 40%. This mirrors the pattern reported for PDTB.
  • Future work: extending the scheme to cover attribution.

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: Discourse analysis

Cited by

32 papers of our library cite it (of the 140673 whose bibliographies we read); 9 of them have a page here, 23 not yet.

Cite

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

@inproceedings{oza2009hindi,
  title = {{The Hindi Discourse Relation Bank}},
  author = {Oza, Umangi and Prasad, Rashmi and Kolachina, Sudheer and Sharma, Dipti Misra and Joshi, Aravind},
  booktitle = {Proceedings of the Third Linguistic Annotation Workshop (LAW III)},
  year = {2009},
  doi = {10.3115/1698381.1698410},
  url = {https://doi.org/10.3115/1698381.1698410}
}

References

7 references, 4 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. Proc. of IJCNLP-2008. find
  2. Lucie Mladová, Šárka Zikánová and Eva Hajičová. 2008. From Sentence to Discourse: Building an Annotation Scheme for Discourse Based on Prague Dependency Treebank. Proc. of LREC-2008. find
  3. Rashmi Prasad, Nikhil Dinesh, Alan Lee, Eleni Miltsakaki, Livio Robaldo, Aravind Joshi, and Bonnie Webber. 2008. The Penn Discourse TreeBank 2.0. find our card
  4. Rashmi Prasad, Samar Husain, Dipti Mishra Sharma, and Aravind Joshi. 2008. Towards an Annotated Corpus of Discourse Relations in Hindi. Proc. of IJCNLP-2008. find our card
  5. Eve Sweetser.1990. From etymology to pragmatics: Metaphorical and cultural aspects of semantic structure . Cambridge University Press. find
  6. Nianwen Xue. 2005. Annotating Discourse Connectives in the Chinese Treebank. Proc. of the ACL Workshop on Frontiers in Corpus Annotation II: Pie in the Sky. find our card
  7. Deniz Zeyrek and Bonnie Webber. 2008. A Discourse Resource for Turkish: Annotating Discourse Connectives in the METU Corpus. Proc. of IJCNLP-2008. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/oza-2009-hindi-discourse-relation-bank/