| CPC G06F 40/40 (2020.01) [G06F 3/0482 (2013.01); G06F 40/211 (2020.01); G06F 40/284 (2020.01)] | 20 Claims |
|
1. A computer-implemented method comprising:
transmitting, by a machine learning service of a multi-tenant cloud provider network, data for a user interface to be presented to a user, the user interface including one or more portions allowing the user to provide user inputs for an explainability configuration, wherein the explainability configuration is to control execution of a user-configurable explanation pipeline for a machine learning explainability analysis to analyze classification inferences generated via a machine learning model for a batch of input text elements, wherein the classification inferences each indicate a sentiment of a corresponding input text element, wherein the user interface allows the user to indicate whether to use a segmenter provided by the machine learning service or a custom segmentation engine within the explanation pipeline;
receiving, at the machine learning service, data for the explainability configuration, wherein the explainability configuration identifies a user-selected custom segmentation engine or code for the custom segmentation engine, wherein the custom segmentation engine is to be used for segmenting text during the machine learning explainability analysis;
executing the user-configurable explanation pipeline as part of the explainability analysis for the batch of input text elements, wherein the execution for an input text from the batch includes:
segmenting the input text using the custom segmentation engine to yield candidate segments,
generating a mask corresponding to the candidate segments, and
executing an explanation algorithm based at least on use of the candidate segments and the mask to yield a result, wherein the result indicates sentiment scores associated with ones of the candidate segments; and
transmitting the result to be presented to the user.
|