Using Properties to Compare Both Words and Clauses

摘要

There are many applications that use the semantic similarity of words to compare input to stored data, such as conversational agents. When a human thinks of a word they consider a meaning which has associated attributes and properties. Clauses allow words to combine together to form a single meaning. This fact allows words and clauses to be treated in the same way as a collection of properties expressed as words. In this paper, a novel mechanism is proposed to allow the similarity between two ideas expressed as properties to be found. An investigation using definitions as the source of information showed that the new method gives strong correlation.