• When we talk about keyword extraction, we tend to assume a one-way model where input sentences are given and keywords are extracted from them.

But didn’t that model simply precede it because it was easier to study?

The Case for a Chat-Based Language Support System - Listening Chat System

  • User input
  • Extract keywords and ask questions about them
  • The user also enters
  • At this time, keyword extraction for the second input should not be independent of the first input
  • I dare to say important things twice.
  • If you think a keyword is not captured, put it in brackets, etc.
  • What would make users happy if they could get?
    • handle on a vague concept.
    • Discover unexpected connections to other documents for something you had vaguely thought about

A system that fosters understanding in the process of network configuration - What is understanding? - Understanding by System

  • There are various patterns depending on the original data.
    • Multiple books by one author
      • Obtain a mental model of authorship.
    • Multiple books on a subject
      • A few authors
      • What is common and what is different?
    • Submissions by many authors
      • For example, in corporate activities, a bunch of complaints from a large number of users or a bunch of support inquiries.
      • I want to find out the few but important opinions
      • We don’t have the resources to deal with everything, so we want to solve as many customer problems as possible.

This page is auto-translated from /nishio/対話的なキーワード生成 using DeepL. If you looks something interesting but the auto-translated English is not good enough to understand it, feel free to let me know at @nishio_en. I’m very happy to spread my thought to non-Japanese readers.