| CPC G06F 16/24578 (2019.01) [G06F 7/14 (2013.01); G06F 16/212 (2019.01); G06F 16/9532 (2019.01); G06F 16/9538 (2019.01)] | 13 Claims |
|
1. A method, comprising:
generating, by a large language model (LLM), a plurality of search queries from a user input term obtained from a user application;
executing, by the LLM, the plurality of search queries to obtain a plurality of verified search results regarding a plurality of candidates;
generating a plurality of web presence schemas corresponding to the plurality of candidates from the plurality of verified search results;
web-scraping a plurality of websites of a subset of candidates selected from the plurality of candidates to obtain a set of corresponding web-scraping payloads, wherein the web-scraping is a recursive operation;
merging the corresponding web-scraping payloads with respective web presence schemas corresponding to respective candidates of the subset of candidates to obtain a plurality of updated web presence schemas, wherein merging comprises:
for a first candidate of the subset of candidates:
obtaining a set of attribute values for a first entity attribute of a first web presence schema corresponding to the candidate, wherein:
a first attribute value of the set of attribute values is obtained from a first web-scraping payload of the corresponding web-scraping payloads corresponding to the first candidate, and the first web-scraping payload has a highest merge priority,
a second attribute value of the set of attribute values is obtained from the first web presence schema, and the first web presence schema has a second highest merge priority, less than the highest merge priority, and
a third attribute value of the set of attribute values is obtained from a highest confidence result of the set of highest confidence results,
wherein the highest confidence result corresponds to the first candidate, the highest confidence result has a confidence score that is higher than confidence scores of other highest confidence results corresponding to the first candidate, and the highest confidence result has a third highest merge priority, less than the second highest merge priority,
obtaining a first confidence score of the first attribute value, a second confidence score of the second attribute value, and a third confidence score of the third attribute value,
responsive to the first attribute value being a non-null value, and the first confidence score being higher than an attribute confidence threshold of the first entity attribute, selecting the first attribute value as a final attribute value for the first entity attribute,
responsive to the first attribute value being a null value, and the second attribute value being a non-null value, and the second confidence score being higher than the attribute confidence threshold of the first entity attribute, selecting the second attribute value as the final attribute value for the first entity attribute, and
responsive to the second attribute value being a null value and the third attribute value being a non-null value, and the third confidence score being higher than the attribute confidence threshold of the first entity attribute, selecting the third attribute value as the final attribute value for the first entity attribute; and
presenting the plurality of updated web presence schemas of the subset of candidates in the user application.
|