Research default probability for adding missing characters to unigram tokenizers
Open
Nobody has claimed this yet.
pipeline 3: preprocess
research
- Dominant language
- Python
- Stars
- 39
- Forks
- 7
- Avg merge
- 1d 9h
- Merged PRs (30d)
- 5
Description
Current approach is assigning every new character a very low probability ( $log(p) = -18$ ).
- Random distribution with mean matching the existing tokenizer and standard deviation 0
- High probability relative to existing tokens
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
No files, tests, or entry points are named. Review the unigram tokenizer implementation and its handling of missing characters, then compare the fixed log(p) = -18 approach with the proposed probability distributions; done means selecting and documenting a defensible default probability strategy.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100