Information retrieval II: The chaining technique for associative sentence retrieval

Most METHODS of information retrieval attempt to extract, from a file, only single items which completely satisfy a search request. When the input question is highly restrictive, no item in the entire file will be able to match the question, but a satisfactory answer can often be formed by a connected sequence of items, each corresponding to a distinct portion of the search request. This method reduces to the common IR algorithms as the question is broadened. Although the concepts that will be described in this paper could be applied to any general information retrieval task (e.g., document retrieval), we will discuss its application as a method of sentence retrieval which might be part of a general question-answering program.