Establishing industry ground truth

A computer program product comprising:

• a computer readable storage medium readable by one or more processor and storing instructions for execution by the one or more processor for performing a method for building a ground truth table, comprising:
• determining a category of a domain having content documents, wherein the category is an industry class for a plurality of products, and wherein the content documents of the web domain respectively correspond to the plurality of products available from the web domain, the determining comprising:
• harvesting words of the content documents of the domain;
• and
• ascertaining the category of the domain responsive to analyzing the words from the harvesting by use of a natural language analysis tool;
• selecting one or more similar domains associated with the category of the domain, wherein the one or more similar domains preexist the web domain as being operational in the category and being coupled to the web domain via a communication network;
• identifying one or more search terms frequently used in the similar domains from the selecting;
• discovering one or more answer relevant to a search query amongst the content documents of the web domain, each of the one or more answer comprising a product identification code that uniquely identifies a product amongst the plurality of products of the web domain, by performing a first domain search in the domain by use of a search query analogous to a first search term of the one or more search terms from the identifying;
• creating a first ground truth in the ground truth table, wherein the first ground truth including:
• the search query, one or more answer to the search query, and one or more relevance score respectively corresponding to each of the one or more answer, wherein the search query is analogous to the first search term as represented in the content documents of the domain, wherein the one or more answer is discovered from the content documents of the domain as being relevant to the search query, wherein a relevance score corresponding to an answer of the one or more answer indicates how relevant the answer is to the search query within the category of the web domain,
• wherein the relevance score is selected from a predefined number of levels, and wherein the relevance score dictates a manner of presenting the answer within the first ground truth of the ground table;
• and;
• producing the ground truth table for later searches of the domain such that the later searches yield relevant answers as being presented in the ground truth table.

A computer implemented method comprising:

• determining, by one or more processor of a computer, an industry of a launching e-commerce domain having content documents, wherein the industry indicates a class for a plurality of products offered by the launching e-commerce domain, and wherein each of the products available from the launching e-commerce domain is represented in one of the content documents of the launching e-commerce domain;
• selecting one or more benchmarking e-commerce domains associated with the industry of the launching e-commerce domain, wherein the one or more benchmarking e-commerce domains are preexisting and operational in a digital communication network coupled to the launching e-commerce domain;
• identifying one or more frequent search terms of the industry that are used extensively by users in the one or more benchmarking e-commerce domains from the selecting;
• discovering, from the content documents of the launching e-commerce domain, one or more answers associated with a frequent search term amongst the one of the one or more frequent search terms in the one or more benchmarking e-commerce domains from the identifying, by performing a domain search internal to the launching e-commerce domain with a search query based on the frequent search term, wherein each of the one or more answer comprising a product identification code that uniquely identifies a product amongst the plurality of products of the launching e-commerce domain, wherein each of the one or more answer is associated with a relevance score indicating how relevant the product identified in each answer corresponding to the relevance score is to the frequent search term subject to the search query, and wherein each relevance score is selected from a predefined number of levels;
• creating a ground truth entry in a ground truth table of the launching e-commerce domain, the ground truth including:
• the search query, the one or more answer to the search query, and the one or more relevance score respectively corresponding to the one or more answer to the search query, wherein the creating further comprising:
• sorting the one or more answers in the ground truth entry in a descending order of respective relevance scores of the one or more answer to the search query, from one or more most relevant answer corresponding to a highest level amongst the predefined number of levels for the relevance score;
• and
• discarding, from the one or more answer to the search query in the ground truth entry, one or more least relevant answer corresponding to a lowest level amongst the predefined number of levels for the relevance score;
• iterating the discovering and the creating for each frequent search term in the one or more frequent search terms identified in the one or more benchmarking e-commerce domains, resulting in a plurality of ground truth entries respectively corresponding to each of the one or more frequent search terms;
• and
• producing the ground truth table for later searches by users of the launching e-commerce domain to thereby yield search results on the later searches by the users based on the plurality of the ground truth entries stored in the ground truth table of the launching e-commerce domain.