UW-CSE at SemEval-2016 Task 10: Detecting Multiword Expressions and Supersenses using Double-Chained Conditional Random Fields

Proceedings of SemEval-2016, pages 931–936, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics UW-CSE at SemEval-2016 Task 10: Detecting Multiword Expressions and Supersenses using Double-Chained Conditional Random Fields Mohammad Javad Hosseini Noah A. Smith Su-In Lee Computer Science and Engineering University of Washington Seattle, WA 98195, USA {hosseini,nasmith,suinlee}@cs.washington.edu Abstract We describe our entry to SemEval 2016 Task 10: Detecting Minimal Semantic Units and their Meanings. Our approach uses a discrim- inative first-order sequence model similar to Schneider and Smith (2015). The chief nov- elty in our approach is a factorization of the la- bels into multiword expression and supersense labels, and restricting first-order dependencies within these two parts. Our submitted models achieved first place in the closed competition (CRF) and second place in the open competi- tion (2-CRF). 1 Introduction Schneider and Smith (2015) argued that the prob- lems of segmenting a piece of text into minimal se- mantic units, and of labeling those units with seman- tic classes (e.g., supersenses), are intimately con- nected. We propose to use a double-chained conditional random field (which we refer to as “2-CRF,” an example of a factorial CRF; §3.4) for joint multi- word expression identification and supersense tag- ging. Like other CRFs, 2-CRF is a feature-rich probabilistic model that can represent probabilistic dependencies between features and labels and be- tween the labels of the consecutive words. The 2- CRF models local dependencies between MWE and supersense sequences with two parallel chains of la- bels, restricting direct interaction between the two to local, single-word positions. Label constraints on tag bigrams ensure a globally consistent tagging. Our experiments show that 2-CRF outperforms a zero-order baseline, the structured perceptron used by Schneider and Smith (2015), and a conventional CRF (§4). For SemEval 2016 Task 10: Detecting Minimal Semantic Units and their Meanings, we submitted a CRF for the closed condition and a 2- CRF (incompletely trained) for the open condition, achieving first and second place, respectively. 2 Task Description For completeness, we briefly review the shared task. The shared task training dataset, called “Detecting Minimal Semantic Units and their Meanings” (DiM- SUM) (Schneider et al., 2016),1 consists of sen- tences with multiword expression (MWE) and su- persense annotations. The data combine and har- monize the STREUSLE 2.1 corpus of web reviews (Schneider and Smith, 2015)2 and Ritter and Low- lands Twitter datasets (Johannsen et al., 2014).3 Similar to prior work (Schneider and Smith, 2015), the annotation for MWEs extends the con- ventional BIO scheme (Ramshaw and Marcus, 1995) to include gappy MWEs with one level of nesting.4 Segmentations are represented using six tags; the lower-case variants indicate that an expres- sion is within another MWE’s gap. • O and o: single word expression • B and b: the first word of a MWE 1https://github.com/dimsum16/ dimsum-data/blob/1.5/README.md 2http://www.cs.cmu.edu/˜ark/LexSem 3https://github.com/coastalcph/ supersense-data-twitter 4Unlike in Schneider and Smith (2015), there is no notion of weak and strong MWEs. 931 • I and i: a word continuing a MWE We call a tag sequence valid if it matches …

Content truncated to 500 words