Speech processing method and apparatus therefor

A non-transitory computer-readable recording medium in which a computer program for performing a method using a computer is recorded, wherein the method comprises:

• converting a user's spoken utterance comprising a voice command into a user utterance text;
• discovering a domain to which the user utterance text belongs and an intent of the user utterance text by performing syntactic analysis or semantic analysis on the user utterance text, and discovering at least one named entity as a result of recognizing a named entity included in the user utterance text;
• determining whether the user's spoken utterance is a complete spoken utterance or an incomplete spoken utterance according to a result of discovering the domain, the intent, and the named entity;
• based on the user's spoken utterance being an incomplete spoken utterance, generating a spoken query utterance that requests information for completing the incomplete spoken utterance as a complete spoken utterance and feeding the spoken query utterance back to the user by:
• determining a slot lacking a named entity among required slots in the user utterance text based on a pre-established state table comprising a required slot in a user command text according to the domain and the intent, and a query text to be requested corresponding to the required slot;
• generating the query text corresponding to the slot lacking a named entity based on the state table;
• and
• converting the generated query text into a spoken query utterance and feeding the spoken query utterance back to the user;
• and
• constructing a complete spoken utterance by receiving a user's spoken response utterance which responds to the spoken query utterance.

A speech processing apparatus, comprising one or more processors configured to:

• convert a user's spoken utterance comprising a voice command into a user utterance text;
• discover a domain to which the user utterance text belongs and an intent of the user utterance text by performing syntactic analysis or semantic analysis on the user utterance text, and discover at least one named entity as a result of recognizing a named entity included in the user utterance text;
• determine whether the user's spoken utterance is a complete spoken utterance or an incomplete spoken utterance according to a result of discovering the domain, the intent, and the named entity;
• based on the user's spoken utterance being an incomplete spoken utterance, generate a spoken query utterance that requests information for completing the incomplete spoken utterance as a complete spoken utterance and feed the generated spoken query utterance back to the user by:
• determining a slot lacking a named entity among required slots in the user utterance text based on a pre-established state table comprising a required slot in a user command text according to the domain and the intent, and a query text to be requested corresponding to the required slot;
• generating the query text corresponding to the slot lacking a named entity based on the state table;
• and
• converting the generated query text into a spoken query utterance, and feed the spoken query utterance back to the user;
• and
• construct a complete spoken utterance by receiving a user's spoken response utterance which responds to the spoken query utterance.