Systems and methods for constructing an artificially diverse corpus of training data samples for training a contextually-biased model for a machine learning-based dialogue system

Abstract

Systems and methods for constructing an artificially diverse corpus of training data includes evaluating a corpus of utterance-based training data samples, identifying a slot replacement candidate; deriving distinct skeleton utterances that include the slot replacement candidate, wherein deriving the distinct skeleton utterances includes replacing slots of each of the plurality of distinct utterance training samples with one of a special token and proper slot classification labels; selecting a subset of the distinct skeleton utterances; converting each of the distinct skeleton utterances of the subset back to distinct utterance training samples while still maintaining the special token at a position of the slot replacement candidate; altering a percentage of the distinct utterance training samples with a distinct randomly-generated slot token value at the position of the slot replacement candidate; and constructing the artificially diverse corpus of training samples based on a collection of the percentage of the distinct utterance training samples.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS[0001]This application is a continuation of U.S. patent application Ser. No. 16 / 907,792, filed 22 Jun. 2020, which claims the benefit of U.S. Provisional Application No. 62 / 870,180, filed 3 Jul. 2019, US Provisional Application No. 62 / 870,156, filed 3 Jul. 2019, U.S. Provisional Application No. 62 / 890,461, filed 22 Aug. 2019, and U.S. Provisional Application No. 62 / 890,296, filed 22 Aug. 2019, which are incorporated herein in their entireties by this reference.GOVERNMENT RIGHTS[0002]The subject matter of the invention may be subject to U.S. Government Rights under National Science Foundation grants: NSF SBIR Phase 1 Grant—1622049 and NSF SBIR Phase 2 Grant—1738441.TECHNICAL FIELD[0003]The inventions herein relate generally to the machine learning and artificially intelligent dialogue systems fields, and more specifically to a new and useful system and method for intelligently classifying unstructured data into a machine learning-based conversat...

AI Technical Summary

Problems solved by technology

While the rules-based approach for implementing a virtual assistant may be useful for addressing pointed or specific queries or commands made by a user, the rigid or finite nature of this approach severely limits a capability of a virtual assistant to address queries or commands from a user that exceed the scope of the finite realm of pointed and / or specific queries or commands that are addressable by the finite set of rules that drive the response operations of the virtual assistant.

That is, the modern virtual assistants implemented via a rules-based approach for categorizing user input and generating responses to users may not fully satisfy queries and commands posed by a user for which there are no predetermined rules to provide a meaningful response or result to the user.

Images