Developed for the purpose of "discovering information"
AI that reproduces "human judgment" with a small amount of teacher data


"KIBIT" is an artificial intelligence engine from Japan that was originally developed by FRONTEO by combining Landscaping, an artificial intelligence-related technology, and behavioral information science.
By combining the "subtlety" of the human mind (KIBI) and the unit of information "bit" (BIT), it means "artificial intelligence that can learn the subtleties of humans."
By reading the meaning of sentences from texts and learning the tacit knowledge and feelings of people, it is possible to reproduce judgments and how to select information on behalf of people.

FRONTEO's original algorithm that supports KIBIT


Think of big data as a collection of human thoughts and activity results
Science that analyzes human thoughts and behaviors from there

Behavioral Information Science



AI that learns tacit knowledge of people

AI that learns tacit knowledge of people

KIBIT can reproduce the "tacit knowledge" possessed by experts and people with good intuition, without relying solely on keywords in text analysis.By using a unique machine learning algorithm that sets it apart from conventional text mining and search techniques, it is possible to analyze in a short time with a small amount of computational resources and a small amount of teacher data.In addition, it is highly evaluated for its high natural language processing technology and practical results, such as the ability to capture similarities and contexts in different expressions and extract the features of sentences.


KIBIT was born in the legal tech field

KIBIT was born in the legal tech field

FRONTEO has developed an artificial intelligence engine called "KIBIT" in the uncompromising task of finding evidence in international proceedings.Finding evidence in a proceeding requires finding evidence from a large amount of textual data in a limited amount of time."KIBIT", which has been cultivated and developed in such legal business, learns "tacit knowledge" such as expert experience and intuition, and realizes prediction of future behavior from analysis of human thoughts.


High-speed, high-precision, short-term implementation is possible with a small amount of teacher data (from dozens of cases)

High-speed, high-precision, short-term implementation is possible with a small amount of teacher data (from dozens of cases)

KIBIT implements Weight Refinement, a unique algorithm that optimizes the importance of words. Weight Refinement makes it possible to capture the characteristics of data that cannot be captured only by calculation based on the amount of transmitted information, and even with a small amount of data, sufficient performance can be achieved, and learning can be performed without building a large-scale server environment. It has the feature.



Evolving KIBIT

The current KIBIT is "artificial intelligence" that operates by being incorporated into FRONTEO applications, but next-generation KIBIT will evolve into an "KIBIT platform" that operates autonomously in response to external requests, independent of the application. It is said.We are also conducting research that enables analysis of non-linguistic data with a multimodal mechanism that simultaneously processes multiple types of different data such as text, images, and voice.Furthermore, we are also advancing the concept of creating an autonomously distributed artificial intelligence that allows artificial intelligence to communicate with each other in an autonomously decentralized manner and proceed with processing, and the analysis range will expand automatically at high speed.

The integrated platform that supports these complex mechanisms will be the future "KIBIT platform" that autonomously processes big data and obtains advanced comprehensive judgments.

Current status: Incubation of artificial intelligence-related business using language processing technology Phase 2: KIBIT is independent of applications and operates autonomously in response to external requests Phase 3: Multiple types of different data such as text, images, and audio Implemented a multi-modal mechanism that processes at the same time Phase XNUMX: Artificial intelligence communicates autonomously with each other to process data and implement mutual learning

The future brought by KIBIT

Currently, artificial intelligence is generally introduced by a data scientist or machine learning engineer who builds a system at the time of introduction, and the cost at the time of introduction is high, so there is a problem that not everyone can easily use artificial intelligence. ..
With the KIBIT platform currently under development, anyone can easily use the know-how possessed by individual data scientists and machine learning engineers, so that the benefits of KIBIT can be delivered to more people in a shorter period of time. Will be.

At FRONTEO, we believe that the development of artificial intelligence can be broadly divided into two stages.The first stage is the increasing use of artificial intelligence and data within existing organizations to dramatically improve productivity.For example, call center artificial intelligence will streamline the analysis of customer feedback, and factory-controlled artificial intelligence will be introduced for on-demand production.

The next stage is the stage where artificial intelligence collaborates beyond existing organizations and data is used more widely.For example, the artificial intelligence of the call center, which is judged to have many complaints about new products, cooperates with the artificial intelligence that manages the production of the factory to reduce the production volume in the short term.
Furthermore, it is expected that such cooperation of artificial intelligence will be carried out not only across existing organizations but also across industrial boundaries, and a complex ecosystem will be built.Also, in this process, it will be generalized to analyze multiple modality such as language, numerical value, and image in an integrated manner.

If KIBIT evolves in the future and the next-generation "KIBIT platform" is created, it will be possible to perform more sophisticated and convenient analysis through modality other than language and mutual learning between artificial intelligences.Please look forward to the next generation of KIBIT that will lead the new future.

ja 日本語
Machine Translation by Google. : close x