摘要 |
PROBLEM TO BE SOLVED: To upgrade precision of a model that assists in a specific action which a retrieving user intends.SOLUTION: A flag appending unit 1 appends a flag of any of an action word, place word, and attribute word to a keyword of a query for retrieval. A queue conversion unit 2 decomposes a queue having as an element URL information of a predetermined click frequency or document information of a predetermined keyword co-occurrence frequency which is obtained through the retrieval, performs dimensional transform, and creates a queue having as an element a feature value of the keyword. A feature extraction unit 3 calculates a feature vector dependent on action type information of the keyword on the basis of a set of keywords to which the flag is appended, the queue having as an element the feature value, and a set of keywords to which action type information representing a type of action is appended in addition to the flag. A prediction model production unit 4 creates a prediction model of an action type, which is intended by a retrieving user and based on the query, through learning using the feature vector as training data.SELECTED DRAWING: Figure 1 |