Adapting Coreference Resolution to Twitter Conversations

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 2454–2460 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 2454 Adapting Coreference Resolution to Twitter Conversations Berfin Aktas¸∗ SFB1287, Research Focus Cognitive Sciences University of Potsdam, Germany berfinaktas@uni-potsdam.de Veronika Solopova∗ Free University of Berlin, Germany solopov97@zedat.fu-berlin.de Annalena Kohnert Department of Language Science and Technology Saarland University, Germany annalenakohnert@gmail.com Manfred Stede SFB1287, Research Focus Cognitive Sciences University of Potsdam, Germany stede@uni-potsdam.de Abstract The performance of standard coreference reso- lution is known to drop significantly on Twit- ter texts. We improve the performance of the (Lee et al., 2018) system, which is origi- nally trained on OntoNotes, by retraining on manually-annotated Twitter conversation data. Further experiments by combining different portions of OntoNotes with Twitter data show that selecting text genres for the training data can beat the mere maximization of training data amount. In addition, we inspect several phenomena such as the role of deictic pro- nouns in conversational data, and present ad- ditional results for variant settings. Our best configuration improves the performance of the ”out of the box” system by 21.6%. 1 Introduction and Related Work Twitter messages present a discourse genre that in- cludes noisy informal language with abbreviations and purposeful typos, use of nonstandard symbols such as # and @ signs, unintended misspellings, etc., which makes them challenging for NLP ap- plications. We are here interested in the task of automated coreference resolution for nominal men- tions in Twitter conversations, i.e., threads of mes- sages that specifically reply to one another. In addition to non-standard words, Twitter conversa- tions also show peculiar phenomena of referring, such as exophoric pointers to non-linguistic content in attached visual media, and mixed pronominal references to the same entity due to the nature of multi-user conversations (Aktas¸ et al., 2018). Thus, tweets are a complicated genre for coreference resolution, but at the same time highly relevant for many applications that seek to extract information or opinions from users’ messages. In this paper, we use a state-of-the-art resolution system built with the OntoNotes corpus (Pradhan ∗* indicates equal contribution. et al., 2007) and experiment with adding annotated Twitter conversations to the training data. Next, we consider the different – spoken and written – genres included in the OntoNotes corpus. We thus conduct experiments with training on different portions, and we show that carefully selecting genre subsets beats the straightforward ”taking as much as possible”. Overall, our best configuration improves the ”out of the box” performance of the system by Lee et al. (2018) on Twitter data by 21.6%. To our knowledge, there is no work specifi- cally on adapting coreference resolution to Twitter, other than the aforementioned study of Aktas¸ et al. (2018), which showed a significant drop in performance when a system with OntoNotes models is applied to Twitter. More generally, one of the few studies on domain adaptation for coreference resolution is (Do et al., 2015), which adapts the Berkeley system (Durrett and Klein, 2013) to narrative stories. Do et …

Content truncated to 500 words