Succeeded in improving KIBIT to process Japanese

Enhanced technology for Asian language processing

FRONTEO Inc.
Masahiro Morimoto, President and CEO
2-12-23 Konan, Minato-ku, Tokyo
(Code number: 2158 TSE Growth)

FRONTEO Co., Ltd. (Headquarters: Minato-ku, Tokyo, President: Masahiro Morimoto, hereinafter referred to as FRONTEO) has succeeded in improving Japanese processing in natural language processing with its proprietary AI engine "KIBIT". I will let you know what I did.

KIBIT was developed in 2012 as an AI that can be operated with very little training data and supports the judgment of experts such as lawyers and legal staff.In e-discovery (electronic discovery) in international litigation, it is equivalent to the judgment of a lawyer involved in investigation within a limited period of time from a huge amount of document data that can reach several terabytes per evidence holder (custodian). It is required to find documents related to the evidence with the accuracy of KIBIT Automator, an AI review tool that uses KIBIT, is used both in Japan and overseas as a legal tech that contributes to the efficiency of this evidence discovery.

Unlike English, Japanese is a language that is not punctuated (notation in which each word is separated by a space), and is accompanied by words that have no meaning on their own, such as particles.Therefore, in Japanese processing by AI, sentences are divided into morphemes such as individual words.^*XNUMXIt requires two techniques: a technique to decompose into three (morphological analysis) and a technique to analyze the morphemes obtained by decomposing.Furthermore, in the latter technique, it is difficult to determine what part of speech a morpheme of a single character, such as ``wa'' or ``ni'', is, which is extracted when writing spaces. The challenge has been to assess the extent to which it is relevant to the judgment of

This time, FRONTEO's research and development team has improved the algorithm in the AI engine "Illumination Forest" that makes up KIBIT so that words consisting of one letter are automatically selected by machine learning.As a result, compared to before the improvement, recall^*XNUMXimproved, resulting in up to a 80% reduction in the number of human-reviewed documents required to discover 7% of the evidence-relevant documents (see figure, using FRONTEO test data).

This technology can also be applied to other non-spaced languages such as Korean and Chinese. One of KIBIT's strengths is that it specializes in Asian languages, which are difficult to process, and the results of this research are expected to further improve the accuracy of a wide range of products equipped with KIBIT.

FRONTEO will continue to promote the sophistication of its own AI solutions, which have strengths in natural language processing, and will strive to develop and improve AI algorithms as a digital forensic discovery vendor that supports experts.

*XNUMX Morpheme: The smallest meaningful linguistic unit
*XNUMX Recall: Recall rate.Percentage of data correctly predicted to be relevant out of all data relevant to the evidence.

■ About KIBIT　URL: https://www.fronteo.com/products/kibit/
"KIBIT" is an artificial intelligence that analyzes text without relying on keywords, using a unique machine learning algorithm that reproduces the "tacit knowledge" possessed by specialists and business experts.With high natural language processing technology, high-precision analysis can be performed in a short time with a small amount of teacher data, contributing to operational efficiency and sophistication in fields such as fraud investigation, intellectual property strategy, skill transfer, and risk prediction. doing.

■ About KIBIT Automator　URL: https://legal.fronteo.com/products/kibit-automator/
"KIBIT Automator" improves the efficiency of document review work in electronic discovery (e-discovery), reduces the burden on workers, and reduces costs, among other discoveries required in the trial proceedings of US civil proceedings. An AI tool developed for the purpose, released in March 2019.We apply the research method used in discovery and utilize AI to examine and analyze large-capacity e-mails and electronic files that are evidence materials.In recent years, it is expected to respond to the short-term information disclosure required of companies.

■ About FRONTEO　URL:https://www.fronteo.com/
FRONTEO is a data analysis company that uses the in-house developed AI engines KIBIT and Concept Encoder, Looca Cross specialized in natural language processing to extract meaningful and significant information from a large amount of textual data and support companies' business. Since its inception in August 2003, we have been working globally with Japan, the United States, Korea, and Taiwan, mainly in the Legaltech business, such as e-discovery (electronic evidence disclosure) and digital forensic surveys, which support company international lawsuits. Based on AI technology developed in this project, we have expanded our business fields to life science, business intelligence, and economic security since 8, and have used AI to transform textual data into knowledge, thus contributing to the resolution of various companies' issues, including support for drug discovery, support for dementia diagnostics, and support for financial, human resources, and business. Listed on TSE Mothers (currently TSE Growth) on June 2014, 2007. In January 6, the company obtained a license to manufacture and sell Class I medical devices (license number: 26B2021X1), and in September of the same year, the company submitted a notification to sell controlled medical devices (notification number: 13MinatoMiSeikikiDai1). Capital: ¥10350 (as of March 9, 3).

* FRONTEO, KIBIT, KIBIT Automator, conceptencoder, and Looca Cross are registered trademarks of FRONTEO in Japan.

<Contact information for the press>
PR Department, FRONTEO, Inc.
Email: pr_contact@fronteo.com

<Contact for legal tech AI business>
FRONTEO Inc. Legal Tech AI Business Headquarters　　　
Email: https://legal.fronteo.com/contact/