Using Semantic Information to Improve Case Retrieval in Case-Based Reasoning Systems
Jyoti Iyer, Pushpak Bhattacharyya · 2005
Abstract. Conventional Case-Based Reasoning (CBR) systems rely on word knowledge to index and search cases from its memory. On being presented with a problem, the Case-Based Reasoning system tries to retrieve a relevant case based on the words that appear in the problem sentence without considering their respective senses. Drawbacks of such systems become more evident in cases where the input is in the form of a sentence in a natural language. Ignoring semantic information in this case may not result in retrieval of desired case or may result in retrieval of an undesired case. In this paper we present a method that tries to improve the precision of retrieval by also taking into account semantic information available to us about the words in the problem sentence. Towards this goal, Universal Networking Language (UNL) is made use of, which provides a semantic representation of natural language text to capture sentence structure. Lexical resource like WordNet is used for finding semantic similarity between two concepts. Using an existing commercial Case-Based Reasoning system as basis for comparison, we demonstrate that considering such semantic information helps in improving case retrieval. 1