Automatic identification of tense and grammatical meaning in context

Abstract

This paper describes the development and evaluation of the Tense and Meaning Identifier (version 2.0). This prototype pedagogic tool categorizes finite verb groups in simple sentences into one of twelve grammatical tenses, and categorizes the verb into classes, such as stative and dynamic, and where applicable subclasses, such as punctual or durative. Using the results of the tense and verb class categorization, the grammatical meaning in context is predicted. Drawing on the Natural Language Toolkit, a program was written to classify finite verbs by their tense and aspect. A tailor-made list of verbs and their associated verb classes and subclasses was created by crawling the web and extracting lists from grammar books. The tailor- made lists are stored in a Python dictionary. A tense-class Python dictionary was also created to look up the corresponding meaning in context. A web app with a submission form was created to enable online submissions and show the tense, verb class and meaning in context for simple declarative sentences. The tense identifier is able to relatively accurately (69% to 100%) identify tenses, but further development is necessary to reduce false positive results. The limitations of this prototype are detailed and suggestions for further work provided.