Transferring information across language understanding model domains

One or more computer storage media having computer-executable instructions embodied thereon that, when executed by a computing processor, perform a method of expanding slot coverage for a target domain-specific natural language understanding (“NLU”) system by transferring information from a source domain-specific natural language understanding model (“source NLU”), the method comprising:

• generating a plurality of intent/entity pairs comprising a phrase pattern and an entity;
• wherein the phrase pattern is mapped to an intent within the source NLU, and wherein the entity is associated with a slot in the target NLU;
• determining a transferability score for pairing the intent with the slot within the target NLU using a classification model that compares the plurality of intent/entity pairs with search records to generate the transferability score;
• and
• upon determining the transferability score satisfies a transferability threshold, transferring the intent to the target NLU for pairing with the slot.

A method of expanding slot coverage for a domain-specific natural language understanding (“NLU”) system, the method comprising:

• receiving a natural language query;
• assigning, using a domain classifier, the query to a specific domain associated with the domain-specific NLU;
• analyzing the query within the domain-specific NLU to determine an intent supported by the domain-specific NLU, the intent having a slot that can be associated with a slot value derived from the query;
• determining that an entity in the query is not recognized by the domain-specific NLU as an accepted slot value;
• validating a transfer of the entity to the domain-specific NLU without retraining the domain-specific NLU;
• and
• upon validation, assigning the entity as the slot value for the intent.