1.4 QUESTION REFORMULATION PROCESSING MODULE. IR SYSTEMS BASED...

2.1.4 Question Reformulation

Processing module. IR systems based on cosine similarity often return documents even if not all keywords are present. Once the “focus” and “question type” are identified, the Information retrieval systems are usually evaluated based on module forms a list of keywords to be passed to the two metrics – precision and recall. Precision refers to the ratio information retrieval component in the document processing of relevant documents returned to the total number of module. The process of extracting keywords could be documents returned. Recall refers to the number of relevant performed with the aid of standard techniques such as named-documents returned out of the total number of relevant entity recognition, stop-word lists, and part-of-speech taggers, documents available in the document collection being etc. searched. In general, the aim for information retrieval systems Other methods of expanding the set of question keywords is to optimize both precision and recall. For question could include using an online lexical resource such as the answering, however, the focus is subtly different. Because a WordNet ontology. The synsets (synonym sets) in WordNet QA system performs post processing on the documents could be used to expand the set of question keywords with returned, the recall of the IR system is significantly more semantically related words that might also occur in documents important than its precision [7]. containing the correct question answer [9].