Project Highlites: Lightweight Open-Source Predictor for any Classifications, Trained with Data of Choice

7 pages•Published: October 19, 2026

Abstract

Subject classification of publications, projects and other entities is ubiquitous in CRIS systems.
This project arose from practical considerations: how could we easily use AI to facilitate the cris administrator's work? This led to an application that can be deployed in various settings: given a brief textual description (e.g., a title and/or abstract of a publication), the application can generate one or more proposals for subject cataloging.
This can be implemented directly in various settings and for different types of classifications. Given sufficient training material, the application can be used with any classification system.
The aim of this article is to demonstrate that the open-source project Highlites is a useful tool for calculating usable classification proposals using one's own limited resources, and in a sufficiently short time.

Keyphrases: artificial intelligence (ai), classification prediction, on premises, open source software

In: Pablo de Castro and Joachim Schöpfel (editors). Proceedings of 17th International Conference on Current Research Information Systems, vol 5, pages 164-170.

Links:	https://easychair.org/publications/paper/B6Ch
	https://doi.org/10.29007/kcs1

BibTeX entry

@inproceedings{CRIS2026:Project_Highlites_Lightweight_Open,
  author    = {Alain Descamps},
  title     = {Project Highlites: Lightweight Open-Source Predictor for any Classifications, Trained with Data of Choice},
  booktitle = {Proceedings of 17th International Conference on Current Research Information Systems},
  editor    = {Pablo de Castro and Joachim Schöpfel},
  series    = {EPiC Series in Technology},
  volume    = {5},
  publisher = {EasyChair},
  bibsource = {EasyChair, https://easychair.org},
  issn      = {2516-2322},
  url       = {/publications/paper/B6Ch},
  doi       = {10.29007/kcs1},
  pages     = {164-170},
  year      = {2026}}

Download PDF Open PDF in browser