Multilingual Aliasing for Auto-Generating Proposition Banks

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3466–3474, Osaka, Japan, December 11-17 2016. Multilingual Aliasing for Auto-Generating Proposition Banks Alan Akbik IBM Research Almaden 650 Harry Road, San Jose CA 95120, USA akbika@us.ibm.com Xinyu Guan Yale University 82-90 Wall Street, New Haven CT 06520, USA xinyu.guan@yale.edu Yunyao Li IBM Research Almaden 650 Harry Road, San Jose CA 95120, USA yunyaoli@us.ibm.com Abstract Semantic Role Labeling (SRL) is the task of identifying the predicate-argument structure in sen- tences with semantic frame and role labels. For the English language, the Proposition Bank pro- vides both a lexicon of all possible semantic frames and large amounts of labeled training data. In order to expand SRL beyond English, previous work investigated automatic approaches based on parallel corpora to automatically generate Proposition Banks for new target languages (TLs). However, this approach heuristically produces the frame lexicon from word alignments, leading to a range of lexicon-level errors and inconsistencies. To address these issues, we propose to manually alias TL verbs to existing English frames. For instance, the German verb drehen may evoke several meanings, including ”turn something” and ”film something”. Accordingly, we alias the former to the frame TURN.01 and the latter to a group of frames that includes FILM.01 and SHOOT.03. We execute a large-scale manual aliasing effort for three target languages and apply the new lexicons to automatically generate large Proposition Banks for Chinese, French and German with manually curated frames. We present a detailed evaluation in which we find that our proposed approach significantly increases the quality and consistency of the generated Proposition Banks. We release these resources to the research community. 1 Introduction Semantic role labeling (SRL) is the task of labeling predicate-argument structure in sentences with shal- low semantic information. The prominent labeling scheme for the English language is the Proposition Bank (Palmer et al., 2005), which provides a lexicon of possible frames for English verbs. Each frame corresponds to one semantic interpretation and comes with frame-specific role labels and descriptions. Over the past decade, large amounts of text data have been annotated based on these guidelines (Palmer et al., 2005; Hovy et al., 2006; Bonial et al., 2014). They enable the training of statistical SRL systems, which have proven useful for downstream applications such as information extraction (IE) (Fader et al., 2011), question answering (QA) (Shen and Lapata, 2007; Maqsud et al., 2014) and machine transla- tion (Lo et al., 2013). However, such manual efforts are known to be highly costly. Possible frames need to be manually determined, their roles individually described, and large volumes of text data annotated accordingly. For this reason, Proposition Banks do not exist for most languages. Annotation projection. Recent research has explored the possibility of using annotation projection to automatically generate Proposition Banks from parallel corpora for new target languages (TL) (Pad´o and Lapata, 2009; Van der Plas et al., 2011; Akbik et al., 2015). This approach requires a large word-aligned corpus of English sentences and their TL translations. …

Content truncated to 500 words