AIC Seminar Series

We consider the problem of checking whether an organization conforms to a body of regulation. Conformance is studied in a runtime verification setting. The regulation is translated to a logic, from which we synthesize monitors. The monitors are evaluated as the state of an organization evolves over time, raising an alarm if a violation is detected. An important challenge to this approach comes from the fact that regulations are commonly expressed in natural language. The translation to logic is difficult. Our goal is to assist in this translation by: (a) the design of logics that let us formalize regulation one sentence at a time, and (b) the use of natural language processing as an aid in the sentential translation. There are many features that are needed in a logic, to accommodate a sentential translation of regulation. We study two such features, motivated by a case study. First, statements in regulation refer to others for conditions or exceptions. Second, sentences in regulation convey legal concepts, e.g., obligation and permission. Obligations and permissions can be nested to convey concepts, such as, rights. We motivate and design a logic to accomodate these two features of regulatory texts. An underlying theme is the importance of the notion of ``saying’’ in such constructs. We then turn to the use of natural language processing to translate a sentence to logic. We study one component of the translation in a supervised learning setting. Linguistic theories have argued for a level of logical form as a prelude to translating a sentence into logic. Logical form encodes a resolution of both predicate-argument structure, and scope ambiguities. We define a restricted kind of logical form, called ``abstract syntax trees’’ (ASTs), based on the logic developed. We describe a modest sized corpus of ASTs annotated on the Food and Drug Administration’s Code of Federal Regulations (FDA CFR), and ongoing experiments on learning to compute ASTs.

Nikhil Dinesh received the B.E. degree in Computer Science from the Regional Engineering College, Trichy, India. He is currently a Ph. D. candidate in Computer Science at the University of Pennsylvania, expecting to graduate in July 2010. His research interests include formal verification, logic, and natural language processing. For his PhD thesis, he focused on the problem of translating regulatory documents to logic, for the applications of conformance checking and access control. He has also been involved in developing the Penn Discourse Treebank, which provides a level of discourse structure and semantics on the Wall Street Journal corpus.

Please arrive at least 10 minutes early in order to sign in and be escorted to the conference room. SRI is located at 333 Ravenswood Avenue in Menlo Park. Visitors may park in the visitors lot in front of Building E, and should follow the instructions by the lobby phone to be escorted to the meeting room. Detailed directions to SRI, as well as maps, are available from the Visiting AIC web page.

©2013 SRI International 333 Ravenswood Avenue, Menlo Park, CA 94025-3493
SRI International is an independent, nonprofit corporation. Privacy policy