US 12,380,356 B2
Identifying noise in verbal feedback using artificial text from non-textual parameters and transfer learning
Isha Aniruddha Oke, Redmond, WA (US); Logan M. Ringer, Mukilteo, WA (US); Vandana Gunupudi, Redmond, WA (US); Yong Hui Ma, Bellevue, WA (US); Prabagaran Santhanakrishnan, Sammammish, WA (US); and Mengtao Sun, Seattle, WA (US)
Assigned to Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed by Microsoft Technology Licensing, LLC, Redmond, WA (US)
Filed on Aug. 17, 2020, as Appl. No. 16/995,350.
Prior Publication US 2022/0051132 A1, Feb. 17, 2022
Int. Cl. G06N 20/00 (2019.01); G06F 16/36 (2019.01); G06F 18/24 (2023.01); G06F 40/237 (2020.01)
CPC G06N 20/00 (2019.01) [G06F 16/36 (2019.01); G06F 18/24 (2023.01); G06F 40/237 (2020.01)] 20 Claims
OG exemplary drawing
 
1. A system comprising:
a processor; and
memory storing computer instructions that, when executed, perform operations to:
receive free-text content corresponding to user satisfaction feedback with performance of a software product or service, wherein the free-text content is naturally spoken or written feedback submitted by a user, and wherein submission of the free-text content causes creation of a data file comprising the free-text content;
in response to submission of the free-text content, retrieve parameter values from the data file, wherein the parameter values are organized in accordance with a schema and comprise an identifier of at least one feature of the software product or service;
generate a free-text corpus based on the user feedback text;
generate an artificial-text corpus by applying a rule to at least one of the parameter values, the rule specifying instructions for converting the parameter values to a set of words, the artificial-text corpus comprising the set of words;
generate first sentence embeddings based on the free-text corpus;
generate second sentence embeddings based on the artificial-text corpus;
generating first combined sentence embeddings by combining the first sentence embeddings and the second sentence embeddings;
generate a classification relating to the free-text content based on the first combined sentence embeddings, the classification indicating whether the software product or service is functioning as expected; and
based on determining the software product or service is not functioning as expected, at least one of providing technical support for the software product or service or updating the software product or service.