Corpus-Driven Thematic Hierarchy Induction

Proceedings of the 22nd Conference on Computational Natural Language Learning (CoNLL 2018), pages 54–64 Brussels, Belgium, October 31 - November 1, 2018. c⃝2018 Association for Computational Linguistics 54 Corpus-driven Thematic Hierarchy Induction Ilia Kuznetsov and Iryna Gurevych Ubiquitous Knowledge Processing Lab (UKP-TUDA) and Research Training Group AIPHES Department of Computer Science Technische Universit¨at Darmstadt http://www.ukp.tu-darmstadt.de/ Abstract Thematic role hierarchy is a linguistic tool used to describe interactions between semantic roles and their syntactic realizations. Despite decades of dedicated research and numerous thematic hierarchy suggestions in the litera- ture, this concept has not been used in NLP so far due to incompatibility and limited scope of existing hierarchies. We introduce an empir- ical framework for thematic hierarchy induc- tion and evaluate several role ranking strate- gies on English and German corpus data. We hypothesize that inducing a thematic hierar- chy is feasible, that a hierarchy can be induced from small amounts of data and that resulting hierarchies apply cross-lingually. We evaluate these assumptions empirically. 1 Introduction Semantic roles are one of the core concepts in NLP, and automatic semantic role labeling (SRL) is a major task with applications in question an- swering (Shen and Lapata, 2007), machine trans- lation (Liu and Gildea, 2010) and information ex- traction (Christensen et al., 2010). The goal of SRL is to label the semantic arguments of a pred- icate (e.g. a verb) with roles from a pre-defined role inventory. Conceptually, role assignment in SRL can be split in two steps: local labeling es- timates the likelihood of a certain semantic argu- ment bearing a certain role; global optimization takes context-dependent role interactions into ac- count and enforces certain theoretically motivated constraints (e.g. “each role must appear only once per predication”). State of the art in SRL is held by the systems based on deep neural networks (Marcheggiani and Titov, 2017; He et al., 2017). While achieving remarkable quality on benchmark datasets, mod- ern systems show a considerable ≈10-point per- formance drop when applied out-of-domain. This issue is aggravated by the fact that deep neural net- works require significant amounts of training data, and SRL annotations are expensive to produce. While local role assignment can be augmented us- ing unlabeled data (e.g. via pre-trained word and character embeddings), context-dependent role in- teraction is an SRL-specific phenomenon and can only be learned from annotated SRL corpora. Aiming to reduce the training data requirements for SRL, we revisit the notion of thematic hierar- chy (TH), a compact delexicalized way to model context-dependent role interactions. Thematic hi- erarchies assume that given a syntactic hierarchy (e.g. subject ≺1 object ≺ oblique) semantic roles can be ranked in a way that higher ranked roles take higher ranked syntactic positions. One exam- ple of phenomena captured by THs is the choice of subject: given a thematic hierarchy Agent ≺ ... ≺ Instrument, an Instrument can only become subject if the Agent is not present, e.g. “[John]Ag broke the window with a [hammer]In” → “A [hammer]In broke the window”. THs have received considerable …

Content truncated to 500 words