githubharald/CTCWordBeamSearch

Connectionist Temporal Classification (CTC) decoder with dictionary and language model.

61
/ 100
Established

When you have an AI model that recognizes sequences like text from an image or spoken words, this tool helps translate the model's raw output into actual words and sentences. It takes the character probabilities from your model and, using a dictionary and optionally a language model, produces a clear, readable text string. This is for AI practitioners, data scientists, or researchers who are building or evaluating sequence recognition systems.

577 stars.

Use this if you need to accurately convert your AI model's character-level predictions into a coherent string of words, especially when dealing with noisy inputs or out-of-vocabulary terms.

Not ideal if your recognition task does not involve sequences of characters that form words, or if you do not have a dictionary or language model to leverage.

automatic-speech-recognition handwritten-text-recognition optical-character-recognition sequence-prediction natural-language-processing
No Package No Dependents
Maintenance 10 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 25 / 25

How are scores calculated?

Stars

577

Forks

159

Language

C++

License

MIT

Last pushed

Jan 31, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/githubharald/CTCWordBeamSearch"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.