Contextual Vocabulary Acquisition of Verbs
Chris Becker · 2005
This paper describes a computational implementation of an algorithm to define an ”unknown” verb based on the information in the context it appears in and any applicable background knowledge that the reader brings to bear. First, it describes some related research that has taken place and how it may be applied to the Contextual Vocabulary Acquisition (CVA) project. Next, it describes the process taken to enhance the current implementation of the CVA verb algorithm. This begins with an analysis of what information best conveys verb meaning, focusing on dictionary definitions as a model. It then moves on to how this information can be retrieved from the context of a verb and how it can be used to categorize the verb or determine its relationship to other ‘known ’ actions. Next it describes the computational implementation of this algorithm and the results after applying it to a sequence of passages represented in the SNePS knowledge representation and reasoning system (Shapiro & Rapaport 1987). Lastly, it outlines some of the remaining issues left for future research in the CVA of verbs. 1 Introducing CVA CVA is defined as the ‘active, deliberate acquisition of word meanings from text by reasoning from contextual clues, prior knowledge, language knowledge, and hypotheses developed from prior encounters with the word, but without external sources of help such as dictionaries or people ’ (Rapaport & Kibby 2002). The goal of the CVA project is twofold: to develop and implement algorithms to define an unknown noun, verb, or adjective from the semantic representation of a sentence, and to develop a curriculum to enhance vocabulary learning. The results of the research in each of these areas is then cycled back to the other to further the development of each goal. 1.1 Previous CVA research Previous research on CVA is made up of a very diverse body of literature including computational implementations and psychological studies. One of the goals of our research is to unify this large disparate body of literature and incorporate elements of both disciplines into our algorithms. Over the past few decades, a number of similar approaches have been developed for computational CVA. Each of these systems has strengths that we are hoping to mirror and weaknesses that we are hoping to improve upon.