Semi-Supervised Semantic Role Labeling: Approaching from an Unsupervised Perspective

Proceedings of COLING 2012: Technical Papers, pages 2635–2652, COLING 2012, Mumbai, December 2012. Semi-Supervised Semantic Role Labeling: Approaching from an Unsupervised Perspective Ivan Titov Alexandre Klementiev Saarland University, Saarbrücken, Germany {titov, aklement}@mmci.uni-saarland.de ABSTRACT Reducing the reliance of semantic role labeling (SRL) methods on human-annotated data has become an active area of research. However, the prior work has largely focused on either (1) looking into ways to improve supervised SRL systems by producing surrogate annotated data and reducing sparsity of lexical features or (2) considering completely unsupervised semantic role induction settings. In this work, we aim to link these two veins of research by studying how unsupervised techniques can be improved by exploiting small amounts of labeled data. We extend a state-of-the-art Bayesian model for unsupervised semantic role induction to better accommodate for annotated sentences. Our semi-supervised method outperforms a strong supervised baseline when only a small amount of labeled data is available. KEYWORDS: semantic role labeling, semi-supervised learning, shallow semantics, Bayesian model. 2635 1 Introduction Shallow representations of meaning, and semantic role labels in particular, have a long history in linguistics (Fillmore, 1968). More recently, with the emergence of large annotated resources such as PropBank (Palmer et al., 2005) and FrameNet (Baker et al., 1998), automatic semantic role labeling (SRL) has attracted a lot of attention (Gildea and Jurafsky, 2002; Carreras and Màrquez, 2005; Surdeanu et al., 2008; Hajiˇc et al., 2009). SRL representations encode the underlying predicate-argument structure of sentences, or, more specifically, for every predicate in a sentence they identify a set of arguments and associate each argument with an underlying semantic role, such as an agent (an initiator or doer of the action) or a patient (an affected entity). SRL representations have many potential applications in NLP and have been shown to benefit question answering (Shen and Lapata, 2007; Kaisser and Webber, 2007), textual entailment (Sammons et al., 2009), machine translation (Wu and Fung, 2009; Liu and Gildea, 2010; Wu et al., 2011; Gao and Vogel, 2011), and dialogue systems (Basili et al., 2009; van der Plas et al., 2009), among others. Most of the current statistical approaches to SRL are supervised, requiring large quantities of human annotated data to estimate model parameters. However, such resources are expensive to create and only available for a small number of languages and domains. Moreover, when moved to a new domain, performance of these models tends to degrade substantially (Pradhan et al., 2008). Scarcity of annotated data has motivated the research into techniques capable of exploiting unlabeled data, that is, semi-supervised and unsupervised learning. The existing semi-supervised approaches to SRL can largely be regarded as extensions to supervised techniques, as they use supervised learning as sub-routines in the estimation process. These include self-training and co-training methods (He and Gildea, 2006b; Lee et al., 2007; Kaljahi and Samad, 2010), mono-lingual and cross-lingual annotation projection (Fürstenau and Lapata, 2009; Pado and Lapata, 2009; van der Plas et al., 2011), and methods which exploit or induce word representations to reduce the sparsity of …

Content truncated to 500 words