Beginning of this page.
Displaying present location on the site.
  1. Home
  2. About NEC
  3. NEC Channel! MiTA TV
  4. NEC's text analysis technology for analyzing vast amounts of text data based on its "meaning" leads the world in accuracy!
Main content starts here.

NEC's text analysis technology for analyzing vast amounts of text data based on its "meaning" leads the world in accuracy!

  • YamamotoBut if Recognizing Textual Entailment technology were used, you could first register words that are prohibited, and extract sentences that include terms with the same meaning as prohibited words from a vast quantity of text data. By doing this, administrators can check reports and emails that contain suspicious sentences without missing any, enabling them to perform risk monitoring efficiently.
    Check only all questionable wording for better efficiency and stricter management.
  • MitaI see!
  • YamamotoIncidentally, monitoring is used for more than just identifying risk. It also leads to the discovery of opportunities.
  • MitaWhat sort of opportunities?
  • YamamotoFor example, analyzing text information posted on SNS or other places could lead to improved sales by identifying purchasing needs that you wouldn't have previously realized, and distributing information or coupons for the corresponding product.
  • MitaIt sounds like the range of uses are only limited by your own imagination.
    What is the other application you are targeting?
  • YamamotoIt is possible to "consolidate*" opinions and demands with the same meaning from customer feedback provided to companies and other organizations.
    * NEC refers to this as "overview comprehension."
  • MitaWhat do you mean by consolidate?
  • YamamotoFor example, suppose you were responsible for a new instant ramen product, and when analyzing survey information there was a range of different types of feedback, such as "the soup is thick," "the noodles have a nice texture," or the "the packaging is innovative." To analyze this customer feedback, you'd first need to categorize these into groups, wouldn't you?
  • MitaThat would be quite a difficult task to start with.
  • YamamotoRecognizing Textual Entailment technology handles this difficult task automatically*. If you feed it a large number of survey answers as text data, it will automatically consolidate those that have the same meaning. For example, "the soup is thick," "the noodles have a nice texture," or "the packaging is innovative." Because representative text that summarizes the group in a readily understandable manner is also shown, you can see at a glance what each group is about.
    * To be exact, this solution uses the Implication Clustering technology that NEC developed based on Recognizing Textual Entailment technology.
  • MizuguchiBy organizing customer feedback into groups like this, we can gauge the number of consenting opinions. This can be useful for discovering issues with products or services.
  • MitaInteresting.
  • YamamotoBy the way, Recognizing Textual Entailment technology can be utilized in any scenario as long as it involves text data. It is possible to apply it to emails, SNS comments, and even call center log notes.
  • MitaSo it can be used in a variety of situations!

Highly-accurate analysis is achieved by taking into account word importance and sentence construction

  • MitaWhat do you think NEC's strengths are?
  • YamamotoFirst of all, NEC's Recognizing Textual Entailment technology was rated best in the world*
    * According to NIST (National Institute of Standards and Technology)
  • MitaWow, best in the world!?
  • YamamotoThat's right. Conventional technology performed analysis based on whether the words in two pieces of text matched. This doesn't allow for analysis when different words are used to express the same meaning, or when the same word is used to express a different meaning. However, NEC's Recognizing Textual Entailment technology performs analysis taking into consideration important words throughout the text, as well as sentence structure such as subject and predicate. This makes it possible to recognize when sentences have a different meaning even though the same words are used.
  • MitaIt seems like a detailed analysis of everything right down to the sentence structure would take a lot of time, though...
  • YamamotoActually, with NEC's Recognizing Textual Entailment technology, the process of finding sentences with the same meaning in around 7 million pieces of text data takes about 0.2 seconds.
  • MitaThat's fast!
  • YamamotoBecause it is equipped with this world-leading processing speed, it can be applied to vast quantities of text data, or "big data" in other words.
  • MitaI see what you mean!
Menu starts here.

NEC Channel! MiTA TV

End of menu.

Top of this page