keyword extraction

• Related: keyphrase extraction.
  • This word is used when we are very conscious of not stopping at one word.
  • I’d like to see “100 Personnel Systems for 100 People” or “Evaluating by Harmony, Generalists are Chosen.”
  • Even the world’s “key phrase extraction” is often subject to restrictions such as “noun sequence” or “adjective*noun+ form” - noun phrase approach
    • It’s not possible to extract key phrases like the above with such a constrained method.
      • Key phrases that allow verbs
      • Active bracketing of verbs
  • Often there are times when you want to use a string of characters that do not appear in a sentence as a key phrase.
    • I want them to be connected by an “information sharing” link when the phrases “information sharing” and “sharing information” are used.
      • I’d like to associate “information sharing” with “information sharing.”
      • Key phrases that do not appear in the sentence

technique

• An approach that does not use linguistic knowledge
  • Simple word frequency. - Need for [stopword
    • Throwing away information on word order.
      • The “General Manager’s Association” issue where the idiom is split.
    • Synonyms are considered different
    • cooccurrence
    • co-location
      • N-grams, etc.
      • intra-window co-occurrence
    • intra-document cooccurrence - concentration (of one’s attention)
  • tf-idf
    • Approach to map real-valued scores whereas the stop word was 0/1.
    • ‘The less frequently it appears in other texts, the more appropriate it is to characterize this text.’
      • IDF part
    • Frequent occurrence as a word, but sometimes an important key phrase in the form of an idiom
  • RAKE
• graph based (e.g. graph)
  • Graph word adjacencies and choose the one with the highest rank.
  • Use PageRank
    • TextRank

---

This page is auto-translated from /nishio/キーワード抽出 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.