METHOD FOR EXTRACTING AN USER INTENT AND APPARATUS FOR PERFORMING THE METHOD

摘要

Disclosed are a method for extracting a user speech intention and an apparatus performing the same, which are capable of improving the performance of prediction of a plurality of speech intentions. The method comprises the steps of: extracting context attributes from a training corpus; generating at least one model for extracting speech intentions by learning a classifier for sequentially classifying context attributes with respect to the extracted context attributes; storing the generated at least one model; extracting at least one context attributes from at least one input user speech; predicting at least one user speech intention using the model generated with respect to the extracted at least one context attributes; and extracting at least one user speech intention based on the predicted at least one user speech intention. Therefore, the method and the apparatus may easily detect a plurality of speech intentions included in a speech by using a speech intention indicator, and may improve the performance of prediction of the speech intentions.