CodecCat is a research-driven character encoding detector focused on low-cost routing, compact byte statistics, calibrated ranking, and reusable inference work.
Current research objective:
- match or exceed strong Python charset detection baselines on held-out real-network data;
- minimize expected compute without sacrificing Top-1 accuracy;
- keep selector and scoring logic small enough for an efficient future native implementation.
Active research branch:
research/charset-normalizer-e2e-baseline