Bright! FRONTEO Official Blog

Bright! FRONTEO Official Blog

Introducing FRONTEO's new FAQ system "KIBIT Find Answer"!

October 2018th, 7
Corporate Communication Department
On July 2018, 7, FRONTEO started offering the automatic FAQ system "KIBIT Find Answer" using the artificial intelligence engine "KIBIT" that was originally developed. 20180717.pdf FAQ system that is being introduced in various scenes for the purpose of improving the efficiency of inquiry work and sharing employee knowledge.What are the features of KIBIT Find Answer released this time? KIBIT-equipped products such as KIBIT Find AnswerBi in charge of salesWorked on research and development of artificial intelligence KIBIT with Mrs. Saito, who belongs to the Genes Solution Headquarters, and KIBIT Find AnActions that led the development of swerWe talked to Mr. Iguchi, who belongs to the Institute of Information Science.
Takumi Saito ※Photoleft Sales representative of the department that handles KIBIT-equipped products (Knowledge Probe, Patent Explorer, Email Auditor, Find Answer) belonging to Solution XNUMX Department of Business Solution Headquarters.We draw out the issues and requests that our customers have and carry out activities to propose optimal solutions.As a sales representative, he also provides side support in the development and introduction of Find Answer.
Shinya Iguchi ※Photoright Belonging to the Research and Development Department, Institute of Behavioral Information Science Involved in the research and development of AI products that utilize the artificial intelligence "KIBIT", which was originally developed, and is in charge of the development of Find Answer.Currently, in the legal tech business, he is working on the development of technology and products that use KIBIT to streamline and automate the browsing and selection of electronic texts based on e-discovery (electronic discovery system).

How KIBIT Find Answer works

—What kind of system is KIBIT Find Answer? Saito:KIBIT is an artificial intelligence specializing in text analysis.By combining the "subtleties" of the human mind (KIBI) and the unit of information, "bits" (BIT), it means "artificial intelligence that can learn the subtleties of the human mind."I'm using this KIBITFind Answer is a solution that applies the features of KIBIT, which reproduces "tacit knowledge" based on human experience and judgment, to problem-solving work using FAQs.KIBIT learns the characteristics of questions entered in natural sentences, extracts appropriate content from the answer data that experts have made in the past, and presents it, saving labor in responding to inquiries from other departments and departments within the company. can.

KIBIT Find Answer answer presentation mechanism

It's been less than a month since it was released, but thankfully the response has been great and we have already received many inquiries. There are customers in various phases, such as "data has been converted but not utilized" and "there is no data and we are responding manually", but the common issue of "finding the necessary information efficiently" is for that purpose. I feel that I am explicitly and potentially seeking a mechanism.I think it is a system that can be used by departments that receive highly specialized questions, such as IT help desk staff, legal departments, and human resources and general affairs.

Read the intent of the question from the natural text KIBIT Find Answer

—What are the features of KIBIT Find Answer? Iguchi:There are various things such as ease of introduction, but from the perspective of development, the biggest feature is that "because you can enter in natural sentences, it is easier to get accurate answers than keyword (word) search". I think.In a general search engine, keywords are entered one after another to narrow down the answers in order to reach the desired answer, but then the hits will gradually disappear and the answers themselves will not come up.In addition, some skill is required on the search side, such as what to set as a keyword. But with Find Answer, you don't have to worry about what to set as a keyword.By email or chatJust like asking a person a question, just enter the text as you come up with it.No matter how long the question text is, KIBIT will capture the characteristics of the text, absorb fluctuations in expression, and pick up questions that are close to the past.The more questions you enter, the more KIBIT will learn its characteristics.It is the opposite of the world where keywords are entered and narrowed down. — What does it mean to capture the characteristics of an interrogative sentence? Iguchi:Let's take a look while actually using Find Answer.For example, when looking up how to apply for a business trip, you would probably enter words such as "business trip," "procedure," and "application" in a general keyword search.But in the case of Find Answer, put in the text that you think of asking a question in chat like this. KIBIT wrote the text of this question as "Business trip/ /Decided/It was/ Of/で す/ /,/In-house/To/In/procedure/,/Application/ Of/Message/The/tell me/Please, morpheme* ¹After looking at everything, think for yourself which word is important and decide for yourself.In this sentence, "I'm sorry" or "please" are commonly used words, aren't they?Regarding such expressions, KIBIT decides that it is "not important" by itself.Therefore, it is not necessary for the questioner to specify the keyword.Then, according to the criteria judged by KIBIT, past questions are scored and the one with the highest score is displayed.  * ¹  Morpheme: The smallest linguistic unit that has meaning. This is the past question and answer that KIBIT actually picked up. The first is the words "in-house" and "procedure" in the text entered by the questioner, and the second is "business trip" and "application", which KIBIT judges to be important and pulls them. Different inputters naturally have different notations and expressions, such as differences between interrogative sentences such as "is it?" And "please tell me" and affirmative sentences, and differences in end-of-sentence expressions such as "is" and "please". It fluctuates. KIBIT understands that as well, grasps the characteristics of the question text, and pulls in similar questions in the past.Let's look at another one.This is an example that does not rely on keywords specified by people. In response to the question "Is there a dress code regulation?", KIBIT gives the highest score to the past question "Is there a dress code regulation?"I think that a general search engine sets "dress code" as a keyword, but in that case, "Is there a dress code?" With a high score of KIBIT will not be hit.However, KIBIT looks at the entire sentence, including the "dress code," and decides whether each morpheme is important or not.Looking at these two sentences, the sentence structure is similar to "Are there any regulations?"In other words, KIBIT gives a high score to this past question by judging that "There is no'dress code', but there are'regulations' and'is there?', So is this similar?" It is.It is unique to KIBIT Find Answer that it is possible to read the intent of a question and give the desired answer with high accuracy.

Artificial intelligence that anyone can experience

— What do you expect from KIBIT Find Answer in the future? Saito:Currently, it is used by customers as an FAQ system, but I think it is essentially an in-house knowledge sharing system.Anyone can easily find out the information assets that exist in the company by accumulating them.I think that is the essence of the value of KIBIT Find Answer.We started by receiving and answering questions like a help desk, but in addition to that, it is a mechanism that can be widely applied to the business of searching for and sharing customer information, so it can meet various needs from the viewpoint of the market I think it is a new product that can respond. Iguchi:For the products up to now, it was necessary to prepare and teach KIBIT what to learn in advance, but when Find Answer was put in to speak the words that came to mind, KIBIT immediately responded. I think we were able to create a highly interactive technology that pulls in related data. I think we have expanded the world where KIBIT can be used more interactively.In the future, this technology will be used in places where artificial intelligence that is conscious of the natural language of human beings and is caring is required, for example, in places where there is interactive interaction such as customer support or robots. I want you to use more. Saito:I think that the interactive interface will continue to permeate in the future.Until now, many KIBIT products have been used by experts, but I think it is a new initiative that everyone, not just specific people, can experience the value of using KIBIT.I would like various people to experience this technology.
Inquiries about KIBIT Find Answer Business Solutions Headquarters TEL: 03-5463-6375 Email: kibit-info@fronteo.com http://www.kibit-platform.com/

Articles in the same category

Commonly viewed articles

ja 日本語
Machine Translation by Google. : close x