Learning Transformation Rules for Semantic Role Labeling

Learning Transformation Rules for Semantic Role Labeling Ken Williams Christopher Dozier Research & Development Thomson Legal and Regulatory Eagan, MN 55123 USA {ken.williams|chris.dozier|andrew.mcculloh}@thomson.com Andrew McCulloh Abstract This paper presents our work on Semantic Role Labeling using a Transformation-Based Error- Driven approach in the style of Eric Brill (Brill, 1995). Our approach achieved an overall F1 score of 43.48 on non-verb annotations. We believe our approach is noteworthy because of its novelty in this area and because it produces short lists of human-understandable transfor- mation rules as its output. 1 Introduction to Transformation-Based Error-Driven Learning For the 2004 Conference on Computational Natural Lan- guage Learning (CoNLL), our team has applied the methodology popularized by Eric Brill for part-of-speech tagging and linguistic parsing (Brill, 1995; Brill, 1993). In this methodology, illustrated in Figure 1, a system learns a sequence of rules that best labels training data. These rules are then used to annotate previously unseen data. According to (Brill, 1995), a Transformation-Based Error-Driven learning application is defined by: 1. The initial annotation scheme 2. The space of allowable transformations 3. The iterative algorithm for choosing a transforma- tion sequence The initial annotation may be extremely simple. For example, in a part-of-speech tagging task, the initial an- notation may assign each token its most likely tag without any regard to context (Brill, 1995). The iterative learning algorithm typically consists of simply searching for a rule that maximizes the increase in some objective function using a greedy hill-climbing Unannotated text Initial annotation Learned annotation Truth annotation Learner Transformation rules Figure 1: Overview of general Transformation-Based Error-Driven learning strategy. For the CoNLL shared task, since participants are evaluated by their F1 scores, it is reasonable to use the F1 score as an objective function. We also imple- mented some extensions to the hill-climbing strategy that we describe in Section 2.3. 2 Experimental Setting In our approach, we used three successive learning stages–the first stage tags the verb region V, the second tags the A0 and A1 arguments, and the third tags all re- maining arguments. The output of each stage becomes the initial annotation for the following stage. Therefore, our system only defines an explicit initial annotation for the verb-tagging phase: for each proposition, we initially tag only the single token containing the verb as V. The search for new transformation templates is ter- minated when no new transformation can be found that would improve the objective function by at least 0.03%. 2.1 Transformation templates For the first stage, transformations are generated from the following eight transformation templates: Lengthen [shorten] the end of region1 V by one token if: a,b) followed by chunk with tag=X c,d) followed by token with POS2=X e,f) followed by chunk with tag=X and token with POS=Y g,h) the verb token’s lemma is X In this formulation, “chunk” refers to the IOB2 chunks, and “clause” refers to the nested clause structure (S re- gions) given as task input. “Lemma” refers to the in- finitive form of the verb, identified in the …

Content truncated to 500 words