A Statistical Tree Annotator and Its Applications

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 1230–1238, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics A Statistical Tree Annotator and Its Applications Xiaoqiang Luo and Bing Zhao IBM T.J. Watson Research Center 1101 Kitchawan Road Yorktown Heights, NY 10598 {xiaoluo,zhaob}@us.ibm.com Abstract In many natural language applications, there is a need to enrich syntactical parse trees. We present a statistical tree annotator augmenting nodes with additional information. The anno- tator is generic and can be applied to a va- riety of applications. We report 3 such ap- plications in this paper: predicting function tags; predicting null elements; and predicting whether a tree constituent is projectable in ma- chine translation. Our function tag prediction system outperformssignificantly published re- sults. 1 Introduction Syntactic parsing has made tremendous progress in the past 2 decades (Magerman, 1994; Ratnaparkhi, 1997; Collins, 1997; Charniak, 2000; Klein and Manning, 2003; Carreras et al., 2008), and accu- rate syntactic parsing is often assumed when devel- oping other natural language applications. On the other hand, there are plenty of language applications where basic syntactic information is insufficient. For instance, in question answering, it is highly desir- able to have the semantic information of a syntactic constituent, e.g., a noun-phrase (NP) is a person or an organization; an adverbial phrase is locative or temporal. As syntactic information has been widely used in machine translation systems (Yamada and Knight, 2001; Xiong et al., 2010; Shen et al., 2008; Chiang, 2010; Shen et al., 2010), an interesting question is to predict whether or not a syntactic con- stituent is projectable1 across a language pair. 1A constituent in the source language is projectable if it can be aligned to a contiguous span in the target language. Such problems can be abstracted as adding addi- tional annotations to an existing tree structure. For example, the English Penn treebank (Marcus et al., 1993) contains function tags and many carry seman- tic information. To add semantic information to the basic syntactic trees, a logical step is to predict these function tags after syntactic parsing. For the prob- lem of predicting projectable syntactic constituent, one can use a sentence alignment tool and syntac- tic trees on source sentences to create training data by annotating a tree node as projectable or not. A generic tree annotator can also open the door of solv- ing other natural language problems so long as the problem can be cast as annotating tree nodes. As one such example, we will present how to predict empty elements for the Chinese language. Some of the above-mentioned problems have been studied before: predicting function tags were studied in (Blaheta and Charniak, 2000; Blaheta, 2003; Lintean and Rus, 2007a), and results of pre- dicting and recovering empty elements can be found in (Dienes et al., 2003; Schmid, 2006; Campbell, 2004). In this work, we will show that these seem- ingly unrelated problems can be treated uniformly as adding annotations to an existing tree structure, which is the first …

Content truncated to 500 words