Computing Logical Form on Regulatory Texts

Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 1202–1212, Edinburgh, Scotland, UK, July 27–31, 2011. c⃝2011 Association for Computational Linguistics Computing Logical Form on Regulatory Texts∗ Nikhil Dinesh Artificial Intelligence Center SRI International Menlo Park, CA - 94025 dinesh@ai.sri.com Aravind Joshi and Insup Lee Department of Computer Science University of Pennsylvania Philadelphia, PA - 19104 {joshi,lee}@seas.upenn.edu Abstract The computation of logical form has been pro- posed as an intermediate step in the translation of sentences to logic. Logical form encodes the resolution of scope ambiguities. In this paper, we describe experiments on a modest- sized corpus of regulation annotated with a novel variant of logical form, called abstract syntax trees (ASTs). The main step in com- puting ASTs is to order scope-taking opera- tors. A learning model for ranking is adapted for this ordering. We design features by study- ing the problem of comparing the scope of one operator to another. The scope comparisons are used to compute ASTs, with an F-score of 90.6% on the set of ordering decisons. 1 Introduction May (1985) argued for a level of logical form as a prelude to translating sentences to logic. Just as a parse tree determines the constituent structure of a sentence, a logical form of a sentence represents one way of resolving scope ambiguities. The level of logical form is an appealing layer of modularity; it allows us to take a step beyond parsing in studying scope phenomenon, and yet, avoid the open problem of fully translating sentences to logic. Data-driven analyses of scope have been of in- terest in psycholinguistics (Kurtzman and MacDon- ald, 1993) and more recently in NLP (Srinivasan and Yates, 2009). The focus has typically been ∗This research was supported in part by ONR MURI N00014-07-1-0907, NSF CNS-1035715, NSF IIS 07-05671, and SRI International. on predicting the preferred scopal ordering of sen- tences with two quantifying determiners, for exam- ple, in the sentence “every kid climbed a tree”. In the related problem of translating database queries to logic, Zettlemoyer and Collins (2009) and Wong and Mooney (2007) consider the scope of adjectives in addition to determiners, for example the scope of “cheapest” in the noun phrase “the cheapest flights from Boston to New York”. To our knowledge, em- pirical studies of scope have been restricted to phe- nomenon between and within noun phrases. In this paper, we describe experiments on a novel annotation of scope phenomenon in regulatory texts – Section 610 of the Food and Drug Administra- tion’s Code of Federal Regulations1 (FDA CFR). Determiners, modals, negation, and verb phrase modifiers are the main scope-taking operators. We have annotated 195 sentences with a variant of log- ical form, called abstract syntax trees (ASTs). Our focus is on the problem of computing the AST, given a (variant of a) parse tree of a sentence. The long term goal of this work is to assist in the translation of regulation to logic, for the application of conformance checking. The problem is to …

Content truncated to 500 words