Research & Development

AI Dialog Technology

Background

Types of dialog systems include question and answer where AI answers user questions using FAQs, chat type that keeps conversations going according to user's utterances, and voice instruction that gives operational guidance to users of smartphones and computers, and practical implementations of these type are already underway. OKI is developing a new technology that incorporates knowledge from experts to enable consultative conversation, a whole new type of dialog with AIs. In face-to-face services, people converse with professionals for special matters such as finding a suitable job, planning for a domicile, and consulting on asset management to receive advice, gain awareness, and realize what they want. Utilizing its proprietary AI Dialog Technology, OKI is developing the technology that allows such services to be realized using artificial intelligence, where AI gives people awareness and brings out true needs and true self inside them. OKI is advancing technological development eyeing the human-machine interface application in connection with automation of consumer support and consultation services.

Features

OKI's AI Dialog Technology using the laddering method (*1) has the following features.

  1. Dialog control function enables counseling-like system speech for deep-probing, paraphrasing, summarizing, and information provisioning.
  2. Intention analysis function captures the intention of the user's speech based on context and interprets the spoken content at the semantic level.
  3. It is not necessary to change the system for each interaction; it operates using set of knowledge for each domain loaded in a declaratively described knowledge data called multidimensional ontology (*2).
  4. User-specific information obtained through the dialog is stored, and referenced during the interaction while the same information is also extended by the inference function for higher-quality information provision.
  5. Sentence generation function generates sentences based on user information obtained during the dialog, and quotes it during the dialog.
  6. Using the intention analysis function, the external search data for each domain is analyzed in advance to clarify the intention, which is then matched at the semantic level with the user intention expressed in the information obtained during the dialog.

With these features, even if the user is unclear about what he/she wants to know or cannot adequately convey his/her wishes, the system will prompt the user in a natural flow of dialog to lead him/her to the desired information.

  • *1 Laddering method is a technique to draw out the needs and values of interviewees by continually asking questions that delve deeper and deeper during interviews.
  • *2 A multidimensional ontology is an ontology extended by OKI. Ontology, in general, is knowledge data on a network describing the definition of concepts represented by words and the relationship between concepts. While only static knowledge can be expressed with conventional ontology, OKI’s multidimensional ontology makes it possible to express knowledge that dynamically changes the dialog based on the user information accumulated during the dialog.

AI Dialog System

AI Dialog System Image

Special Contents

      Contact

      Contact