A user’s interaction with a virtual assistant typically involves spoken requests, queries, and commands which often includes disfluencies. This disclosure describes techniques to automatically correct disfluent queries. Per techniques of this disclosure, a disfluency correction machine learning model is utilized to convert a disfluent query to a corresponding fluent query. Lexical features extracted from the disfluent query are utilized to determine a portion of the query that is removed from the disfluent query to convert it to a fluent query. The model is trained using pairs of queries.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.