Detecting and Finding the True Meaning of Jargons

Abstract

While looking at the online bulletin boards, some obscure description could be found in the threads. In most cases, it is usually because that the authors are not willing to mention the true name of somebody or something directly, and in hence take the place of it with a substitute expression which has the same attributes, such as homophonic word or objects which appear similarly. These words and phrases, which are used irregularly in the context, could be called jargons. In this study, we propose a method to detect and understand jargons by considering the relevance of context words in word-embedding expressions. Specifically, we first calculate the topic correlation-coefficient of each word pair occurring in the text to find the jargon. Then, we combine a set of contextual information to find out a list of candidate words which probably deliver the true meaning of the jargon. By comparing the candidate words and the jargon word from 3 aspects (meaning, pronunciation and shape), the true meaning of the jargon could be determined based on the aggregative similarity. Finally, we have conducted an experiment to examine the usefulness of our approach. Additional Keywords and Phrases: Jargon, Word-embedding expression, Word-vectors, Similarity, Natural processing language.