funNLP: a categorized index of Chinese NLP resources
/funNLP
funNLP is a single repo that indexes hundreds of Chinese NLP tools, datasets, and word lists rather than shipping any code of its own. It's genuinely useful as a starting point when you need a specific lexicon — sensitive words, stopwords, idiom dictionaries — and don't want to search GitHub yourself, but the entries are one-line descriptions with no vetting, so treat it as a directory, not a curated best-of.
What is funNLP?
funNLP is a categorized link collection for Chinese natural language processing, built by fighting41love while learning the field. It groups resources under headings like sensitive-word lists, phone/ID/email extraction patterns, name-gender inference, domain lexicons (legal, medical, financial), pretrained word vectors, and Chinese NLP corpora and datasets, with each entry linking out to the actual repo, paper, or download.
What categories are covered
- ✓Sensitive-word, profanity, and 'reactionary'/violence word lists, plus stopword and negation dictionaries for text filtering
- ✓Regex-based extraction for Chinese phone numbers, ID card numbers, and emails, and name-based gender inference
- ✓Domain lexicons spanning law, medicine, finance, automotive parts, idioms, place names, and historical figures
- ✓Chinese word vectors, BERT/ERNIE resources, and NER, relation-extraction, and keyphrase-extraction tools
- ✓Corpora and datasets: chat logs, rumor data, Baidu QA pairs, dependency-parsing annotations, and speech/ASR sets
- ✓A large ChatGPT/LLM-era section added later — model evaluation, open frameworks like MOSS and Chinese-LLaMA-Alpaca, fine-tuning guides, prompt engineering, and document-QA projects
- ✓Novelty entries too, like a lyric generator built on singer Wang Feng's songs and an auto-couplet (对联) generator
Strengths
- ✓Saves the search-engine legwork for common Chinese NLP building blocks — sensitive words, stopwords, name/phone/ID regexes are ready to copy instead of assembling from scratch
- ✓Broad category coverage: legal, medical, financial, and automotive lexicons sit next to general corpora and pretrained word vectors in one place
- ✓Kept current with the field — the LLM/ChatGPT-era section covers fine-tuning, prompt engineering, and document-QA projects, not just classic NLP
- ✓No signup, no install. funNLP is a README, so any linked resource is one click away
Known limitations and caveats
- △No license is listed on the repo, so it's unclear what you're allowed to do with the aggregated list itself (the linked projects carry their own licenses)
- △It's a flat table of one-line Chinese descriptions with no quality bar — entries range from serious research code to abandoned toy projects, and you have to click through and judge each one yourself
- △Links rot: because it just points at other repos, papers, and blog posts, a share of entries will eventually 404 or point at unmaintained projects
- △Almost every description is in Chinese, so non-Chinese-reading users can use the linked tools but not read the collection's own annotations
Other Chinese NLP resource lists
Frequently asked questions
funNLP is still updated, though on an irregular schedule — the README describes updates as ongoing but not on a fixed timetable, and invites users to watch and fork the repo for changes. It has grown well past its original word-list scope, most recently picking up a large block of ChatGPT and LLM-related resources.
funNLP doesn't document a formal contribution process — no CONTRIBUTING guide is called out in the README. In practice it works through GitHub's normal fork-and-pull-request flow, and the maintainer explicitly welcomes people to watch and fork the repository, but there's no described review process for suggested additions.
funNLP is built mainly around Chinese-language NLP — sensitive-word lists, name and phone-number extraction, and Chinese corpora make up most of the entries. It does include some English-only tools too, such as an English profanity list and English NLP data-augmentation resources, plus a large added section of English-language LLM and ChatGPT tooling.
funNLP itself is a link index, not a code library — the repo is a long README table pointing to other GitHub projects, datasets, papers, and blog posts. A handful of entries are downloadable data files hosted directly in the repo, but there's no package here to install or import.
funNLP doesn't rank or star-sort the tools it links to — entries are grouped by category (lexicons, corpora, pretrained models, LLM frameworks) rather than by popularity. To gauge how widely used a linked project is, you'd need to open that project's own GitHub page and check its stats there.
Best use cases
- •Bootstrapping a Chinese text-filtering pipeline: pull the sensitive-word, profanity, and stopword lists instead of writing your own
- •Finding a starting regex or dictionary for a narrow extraction task — Chinese phone numbers, ID cards, or name-based gender guesses
- •Scoping a domain-specific NLP project (legal, medical, financial) by seeing which lexicons and datasets already exist in Chinese
- •Catching up on the Chinese LLM/ChatGPT ecosystem — the added section links open frameworks, fine-tuning guides, and evaluation benchmarks in one place
Who should try it — and who should skip
Worth bookmarking if you build Chinese-language NLP features regularly and want a single jumping-off point instead of five separate searches — the lexicons and regex patterns alone can save an afternoon. Skip it if you want vetted tools, not links. It's also not installable — funNLP is a README, so you'll still be evaluating and integrating each linked project yourself, and it won't help if you only need English-language NLP resources.
Is funNLP worth your time?
ChatGPT, Claude and Perplexity can all read this page. Ask one of them what it makes of funNLP.