bazingagin/npc_gzip
A parameter-free text classification method that uses data compressors (gzip, bz2, lzma) to classify text without any model training.
Not currently ranked — collecting fresh signals.
star history
This repository implements a text classification approach that leverages compression algorithms to measure the similarity between text documents for classification. The method works by compressing concatenated text pairs and comparing compressed sizes as a distance metric, eliminating the need for trained ML parameters. It supports multiple datasets (AG_NEWS, DBpedia, YahooAnswers, etc.) and compressor options (gzip, lzma, bz2).
Frequently asked
- What is bazingagin/npc_gzip?
- A parameter-free text classification method that uses data compressors (gzip, bz2, lzma) to classify text without any model training.
- Is npc_gzip open source?
- Yes — bazingagin/npc_gzip is open source, released under the MIT license.
- What language is npc_gzip written in?
- bazingagin/npc_gzip is primarily written in Python.
- How popular is npc_gzip?
- bazingagin/npc_gzip has 1.8k stars on GitHub.
- Where can I find npc_gzip?
- bazingagin/npc_gzip is on GitHub at https://github.com/bazingagin/npc_gzip.