Search method and apparatus and storage medium

A search apparatus for searching among a plurality of conversation history documents, each of which includes an inquiry from a customer and a response to the inquiry, for the conversation history documents whose content of the inquiry is similar to an inquiry text representing content of a new inquiry from the customer,

• the search apparatus comprising:
• a conversation history document classification unit that classifies the plurality of conversation history documents into a plurality of clusters in accordance with contents;
• a representative document extraction unit that extracts a conversation history document from each of the clusters as a representative document by using a similarity score among a plurality of conversation history documents in each of the clusters;
• and
• a document search unit that searches for first conversation history documents having a relation with a first inquiry and presents, as a result of the search, a first representative document for a first cluster, to which the first conversation history documents detected by the search belong,
• wherein the conversation history document classification unit includes:
• a conversation history document relation calculation unit that calculates a relation between the conversation history documents;
• a relation structure creation unit that creates a relation structure of the conversation history documents on the basis of the calculated relation between the conversation history documents;
• and
• a conversation history document classification unit that classifies the plurality of conversation history documents into the plurality of clusters on the basis of the created relation structure of the conversation history documents.

A storage medium storing a program to be executed by a search apparatus for searching among a plurality of conversation history documents, each of which includes an inquiry from a customer and a response to the inquiry, for the conversation history documents whose content of the inquiry is similar to an inquiry text representing content of a new inquiry from the customer,

• wherein the program causes the search apparatus to execute processing including:
• a first step of classifying the plurality of conversation history documents into a plurality of clusters in accordance with contents;
• a second step of extracting a conversation history document from each of the clusters as a representative document by using a similarity score among a plurality of conversation history documents in each of the clusters;
• and
• a third step of searching for first conversation history documents having a relation with a first inquiry and presenting, as a result of the search, a first representative document for a first cluster, to which the first conversation history documents detected by the search belong,
• wherein the first step includes:
• a fourth step of calculating a relation between the conversation history documents;
• a fifth step of creating a relation structure of the conversation history documents on the basis of the calculated relation between the conversation history documents;
• and
• a sixth step of classifying the plurality of conversation history documents into the plurality of clusters on the basis of the created relation structure of the conversation history documents.