Key-Word Based Query Recognition in a Speech Corpus by Using Artificial Neural Networks

Information Retrieval deals with the easy access to the information based on the user's request, which will be presented in the form of a query. A dialog system that understands spoken natural language queries asks for further information if necessary and produces an answer to the speaker's query. Most of the research works in Information Extraction focus only on written language processing, in which a few are devoted to the study of Spoken Language Information Extraction. This paper discusses a novel technique for recognition of the isolated question words from Malayalam (one of the south Indian languages) speech query. We have created and analyzed a database consisting of 500 isolated question words. Fast Fourier Transform (FFT) and Discrete Cosine transform (DCT) is used for the feature extraction purpose and Artificial Neural Network (ANN) is used for classification and recognition. A recognition accuracy of 85% could be achieved from this experiment.