Introducing a New Algorithm in the AI Review Tool “KIBIT Automator”:Achieving Human-like sensitivity, Unparalleled Accuracy and Explanatory Powerin Judgment

NEW YORK,June 6, 2023 – FRONTEO (TSE: 2158) announces that it has implemented a new algorithm (patent pending) in its proprietary AI review tool "KIBIT Automator.” This innovative algorithm is inspired by the superior intuitive judgment capability of humans to swiftly recognize and assess color tints. By replicating this innate ability, the new algorithm aims to enhance the efficiency and effectiveness of the document review process.

In recent years, the rapid development of the digital society has made the use of IT a routine activity, and the amount of electronic data managed by companies continues to grow. In e-discovery (electronic discovery) and digital forensics for international litigation, reviewing this vast amount of data to find relevant documents is a difficult task. This process requires both speed and quality assurance within strict time constraints. The quality of the review process varies depending on the skill of the reviewer, while the time and cost of the review process increases with the volume of data being reviewed.

KIBIT Automator is an AI tool that contributes to significantly improving the efficiency and quality of document review tasks. The tool incorporates two types of algorithms: one that calculates the relevance of words in a document by assigning them a "weight" and another that accurately predicts the relative accuracy of identifying relevant documents prior to the start of the review process. These integrated functions have demonstrated their effectiveness in maintaining quality, improving efficiency, and significantly reducing the time and costs associated with the extensive review work conducted by companies and attorneys involved in litigation.

The new algorithm combines the advantages of the two previous algorithms with greater accuracy and explanatory capabilities. In the document review process, humans possess the remarkable ability to quickly and accurately assess the relevance of documents. Inspired by the fact that this is similar to the "color bath effect," which enables us to unconsciously and rapidly recognize specific targets, we developed a cognitive model that perceives words in a document as colors and interprets nuances (the frequency and combination of word occurrences) across the entire document as shades of color. The algorithm was designed to replicate the intuitive judgments made by humans in a similar fashion.

With the implementation of the new algorithm, KIBIT Automator is now able to indicate "negative" relevance in addition to the previous "positive" evaluation of word weights. In addition, the tool can calculate not only the overall document score in terms of relevance, but also analyze the distribution of scores within the document. This allows for the identification and highlighting of high-scoring areas which aids in explaining the AI’s judgment – a task typically considered challenging. In-house verification using the peacetime auditing tool "KIBIT Eye", which is already equipped with this algorithm, a comparison of data extraction rates for auditing purposes at a 20% viewing rate revealed an extraction rate of 82.5% for the conventional system, whereas KIBIT Eye achieved an extraction rate of 99.5%. The new algorithm is more versatile and can be used for a wide range of purposes. Since the new algorithm is highly versatile, we believe that it can be deployed in a variety of FRONTEO products in the future to improve their performance and functions.