Showa Denko K.K. (SDK; Tokyo; www.sdk.co.jp) has introduced an artificial intelligence (AI) based search system to look efficiently for information useful for capital investment evaluation from the company's knowledge databases. This new search engine uses an AI system named "KIBIT," which was developed by FRONTEO, Inc., and searches the company's document database for knowledge useful for its internal examination and screening of investment plans including those for production equipment. SDK started operation of the new system at the end of this January. The new AI-based search system to scrutinize the company's database, aiming to make better use of knowledge about safe and stable operation of production facilities accumulated through judgments on investment plans in the past. This database includes information on more than 2,000 cases of investment in the past 20 years. With conventional search engines, it difficult to find appropriate cases to refer from document files recorded on multiple databases.
KIBIT is an AI that simulates "tacit knowledge" held by experts and skilled workers. In order to search databases for cases appropriate to refer, this AI looks up not only key words but also structure of writing and the line of thought in documents on databases, including those in documents contained in attached files. SDK decided to introduce KIBIT-based search engine because this system enables them to extract cases of investment similar to newly suggested investment plans under screening not reliant on examiners' experiences. In a trial run of the new system, SDK confirmed that the KIBIT-based system searched document databases for similar cases of investment and judged degrees of similarity within almost one tenth of the time needed by conventional search systems. In addition, the KIBIT-based system enables SDK to pick up many similar cases simultaneously, and make the most of knowledge and know-how about facility safety measures. SDK will put this new search system into regular use in its process to judge appropriateness of capital investment plans, and consider to extend the use of the new search system to cover similar cases of accidents and abnormalities in production fields.