AUTOMATING REVIEW OF FORMS FOR INTERNATIONAL TRADE TRANSACTIONS: A NATURAL LANGUAGE PROCESSING APPROACH

Abstract

A major challenge in Office Automation is one of automating routine jobs that involve large-scale processing of ill-formed natural language data. Such data are often present in documents such as forms where it is necessary and/or practical to allow latitude in how the forms may be filled. In this paper, we describe a computational model designed to process free-form textual data in application forms for Letters of Credit (LC), which represent a common vehicle for initiating international trade transactions. The model is based on a variation of the case-frame or thematic-role frame instantiation methods. We describe the implementation of the model, report empirical results with real LC applications, and indicate directions we are currently pursuing to improve its performance.