NEC's text analysis technology for analyzing vast amounts of text data based on its "meaning" leads the world in accuracy!
Enormous amounts of text data (big data) such as emails and SNS posts are transmitted during the course of our daily lives. There is growing demand for the analysis and use of this text data in business activities. NEC actually began undertaking research into technology called "Recognizing Textual Entailment" from an early stage, and it has apparently won high praise both in Japan and overseas.
It sounds like utilizing this technology enables text data to be applied in a range of situations, such as risk monitoring for business activities, and analyzing customer feedback for products or services. We actually spend quite a lot of time reading and tabulating the large number of opinions and requests regarding the MiTA TV program we are sent via emails or SNS comments. It would be a tremendous help if this technology would alleviate that. As a result, all of us on the program staff are very interested to hear more! I went and asked for more details about this technology, as well as how it can be put to use, among other things!
Interviewee: Kosuke Yamamoto
Mr. Yamamoto is endeavoring to utilize text analysis technologies such as Recognizing Textual Entailment with big data. His current area of responsibility is central and local governments.
Interviewee: Mr. Hironori Mizuguchi
Mr. Mizuguchi has been active in the field of text analysis for over 15 years. He currently spends each day striving toward applying text analysis technology such as Recognizing Textual Entailment to big data. He is responsible for private enterprises.
The basic concept is to determine whether or not two sentences have the same meaning
- MitaHello, Mr. Yamamoto and Mr. Mizuguchi. Getting straight to the point, can you tell us what kind of technology Recognizing Textual Entailment is?
- YamamotoFirst, have you heard the term "textual entailment recognition" before?
- MitaNo, I haven't.
- YamamotoLet's start the explanation from there. Suppose we have two sentences. The first reads, "Presentation at the client's premises." The second reads, "Demonstration at the customer's office." Although these two sentences are written differently, what happens when you compare their meaning?
- MitaThey are quite similar.
- YamamotoExactly. How about, "I received the client's presentation materials," which also contains many of the same words?
- MitaThe meaning is a little different to the previous sentences.
- YamamotoJust now you compared two sentences, and determined whether they contained the same meaning. Textual entailment recognition is the act of making this decision. The Recognizing Textual Entailment technology we provide performs this on large volumes of text data such as emails, reports, and SNS posts.
- MitaSo despite the tricky-sounding name, this technology basically involves comparing two sentences and determining whether they have the same meaning?
- YamamotoThat's right!
What can this technology be applied to?
- MitaI'm curious to hear what areas Recognizing Textual Entailment technology can be applied to.
- YamamotoThere are a range of applications, but we believe we can satisfy two major needs. This first is the desire to monitor text that passes through email and other systems in large volumes.
- YamamotoFor example, let's assume that the department responsible for risk management is performing monitoring by checking reports and emails to ensure no employees are violating compliance rules. By conventional means, the administrator would need to read screeds of reports and emails to perform this check.
- MitaWe sure do get a ton of email! It takes an enormous amount of time to manage.