Natural Language Understanding Wiki

In CoNLL-2012 coreference resolution task, a speaker identifier is provided for every sentence (for example, in broadcast news). This information has proved valuable and used in many feature sets throughout the literature, sometimes used as the "only feature for resolving pronouns" (Moosavi and Strube, 2014)[1].

Speaker-related features:

  • Speaker indicator itself (Wiseman et al. 2016[2], Yuan et al. 2012[3])
  • "if a mention has a substring overlap with the current speaker" (Wiseman et al. 2016[2])
  • if "a pronoun [...] is a speaker in the sentence." (Chang et al. 2012[4])
  • if "an antecedent is a speaker in the sentence." (Chang et al. 2012[4])
  • "if an antecedent has a substring overlap with a speaker distinct from the current mention’s speaker" (Wiseman et al. 2016)[2]
  • Comparing two speakers (for two pronouns), e.g. two first-person pronouns with different speaker indicators are regarded as negative evidence (Martschat 2013)[5], similar rules are present in Lee et al. (2013, section 3.3.1)[6], Uryupina&Moschitti (2015)[7], Martschat&Strube (2015)[8] and Moosavi&Strube (2014)[1]
  • "If the speaker of the second person pronoun is talking to the speaker of the first person pronoun"?? (Martschat 2013)[5]
  • "In conversations, <you> can corefer only with the previous speaker." (Lee et al. 2013, section 3.3.1)[6] which is not always correct

Multiple-speaker scenarios are regarded as difficult (Yuan et al. 2012)[3].


  1. 1.0 1.1 Moosavi, N. S., & Strube, M. (2014). Unsupervised Coreference Resolution by Utilizing the Most Informative Relations. Computational Linguistics (COLING), (2013), 644–655.
  2. 2.0 2.1 2.2 Wiseman, S., Rush, A. M., & Shieber, S. M. (2016). Learning Global Features for Coreference Resolution. In NAACL-2016 (pp. 994–1004).
  3. 3.0 3.1 Yuan, B., Chen, Q., Xiang, Y., Wang, X., Ge, L., Liu, Z., … Si, X. (2012). A mixed deterministic model for coreference resolution. In Joint Conference on EMNLP and CoNLL-Shared Task (pp. 76–82).
  4. 4.0 4.1 Chang, K.-W., Samdani, R., Rozovskaya, A., Sammons, M., & Roth, D. (2012). Illinois-Coref: The UI System in the CoNLL-2012 Shared Task. In CoNLL Shared Task (pp. 113–117).
  5. 5.0 5.1 Martschat, S. (2013). Multigraph Clustering for Unsupervised Coreference Resolution. Acl 2013, (2012), 81–88. Retrieved from
  6. 6.0 6.1 Lee, H., Chang, A., Peirsman, Y., Chambers, N., Surdeanu, M., & Jurafsky, D. (2013). Deterministic Coreference Resolution Based on Entity-Centric, Precision-Ranked Rules. Computational Linguistics, 39(4), 885–916.
  7. Uryupina, O., & Moschitti, A. (2015). A State-of-the-Art Mention-Pair Model for Coreference Resolution. SemEval 2015, 289–298.
  8. Martschat, S., & Strube, M. (2015). Latent Structures for Coreference Resolution. Transactions of the Association for Computational Linguistics, 3(0), 405–418. Retrieved from