The main advantages presented in our work on text-representing centroid terms are possibilities to:

  • obtain an exact topical description of a text's content, or even a more general, classifying term;
  • comprise a text's content in a single classifier;
  • calculate centroid terms fast for long texts as well as short queries;
  • compare and determine the similarity of texts even if different wordings (e.g. by different authors) are used;
  • analyse texts according to their natural structure (chapters, sections etc.), or even in a fully hierarchical way starting from the sentence level;
  • consider the sequence of words, which significantly determines the contents and meaning (as well as quality) of a text written;
  • enable a deep learning process similar to processes in the human brain.

Next page

12 February 2018