Signal and Data Processing

In this article a pre-processing method is introduced which is applicable in speech recognized texts retrieval task. We have a text corpus, t generated from a speech recognition system and a query as inputs,  to search queries in these documents and find relevant documents. A basic problem in a typical speech recognized text is some error percentage in recognition. This, results erroneously assigning to irrelevant documents.The idea of this proposed method, is to detect error-prone terms and to find similar words for each term. A parameter is defined which calculates the probability for occurring errors in the error-prone words. To recognize similar words for each specific term, based on a criterion called average detection rate (ADR) and levenshtein distance criterion, some candidates are chosen as the initial similar words set. And then, a conversion probability is defined based on the conversion rate (CR) and the noisy channel model (NCM) and the words with higher probability based on a threshold level are selected as the final similar words. In the retrieval process, these words are considered in the search step in addition to the base word.  Implementation result shows a significant improvement up to 30% of F-measure in information retrieval method with consideration of this pre-processing.