| CPC G06F 40/20 (2020.01) [G06F 16/951 (2019.01)] | 20 Claims |
|
1. A computer-implemented method comprising:
obtaining a record comprising a text narration corresponding to context for the record;
generating, by a natural language processing (NLP) model, an encoded narration based on the text narration;
determining a first similarity index between the encoded narration and each encoded reference topic of a plurality of encoded reference topics that correspond to a plurality of predetermined reference topics, respectively;
determining whether the first similarity index between the encoded narration and each encoded reference topic is equal to or greater than a similarity threshold;
when the first similarity index is equal to or greater than the similarity threshold, adding a respective reference topic of the plurality of predetermined reference topics that corresponds to the encoded reference topic associated with the first similarity index to a first result group;
when the first similarity index is less than the similarity threshold, leaving the respective reference topic out of the first result group; and
classifying the record based on a reference topic included in the first result group that corresponds to the encoded reference topic associated with a greatest similarity index within the first result group,
wherein the computer-implemented method further comprises:
based on the first similarity index being less than the similarity threshold:
determining that the encoded narration corresponds to a non-matchable text narration;
performing web scraping based on the non-matchable text narration, to obtain contextual information contextually similar to the non-matchable text narration;
generating, by the NLP model, encoded contextual information based on the contextual information;
determining a second similarity index between the encoded contextual information and each encoded reference topic of the plurality of encoded reference topics;
comparing the second similarity index between the encoded contextual information and each encoded reference topic to the similarity threshold;
when the second similarity index is equal to or greater than the similarity threshold, determining one or more reference topics among the plurality of predetermined reference topics that corresponds to one or more encoded reference topics associated with the second similarity index, respectively, as matching the non-matchable text narration;
adding the one or more reference topics to a second result group; and
classifying the record corresponding to the non-matchable text narration based on a reference topic that is included in the second result group and corresponds to the encoded reference topic having a greatest similarity index within the second result group.
|