Rakuten, Inc.

Mercari, Inc.

Kenji Sugiki

Takuma Yamaguchi

Teiko Inoue

Taku Yasui

Provided is a product information generation system and the like capable of performing comprehensive determination using information on access to a plurality of pieces of media data from media data representing a product and generating product information. A product information generation system including a storage unit that stores access information used for access to product media data representing a product from a plurality of terminal devices in association with the product media data, a communication unit that receives the product media data transmitted from one terminal device, a feature extraction unit that extracts a feature of the received product media data, a similar media acquisition unit that acquires similar product media data of which a degree of similarity to the received product media data is equal to or higher than a predetermined degree of similarity and the access information associated with the similar product media data from the storage unit on the basis of the extracted feature, and an information generation unit that merges a plurality of pieces of access information of the acquired similar product media data to generate product information indicating the product of the received product media data.

Product information generation system, product information generation program, and product information generation method

A pair search word, which is obtained by forming a pair of a preceding search word and a subsequent search word in the search times, is generated in accordance with an order of the search times, from search words in which intervals between the search times associated with the identical user specifying information are within a predetermined time, with reference to a search word memory means which stores the search words. A first appearance count of a pair search word among generated pair search words is calculated, a second appearance count of a reverse order pair search word obtained by reversing an order of a search time of the specific pair search word is calculated, and, when a magnitude relationship between the first appearance count and the second appearance account satisfies predetermined conditions, the preceding search word and the subsequent search word are stored as a thesaurus.

Growing community of inventors

Tokyo, Japan

Kenji Sugiki