Skip to content

crownpku/Rasa_NLU_Chi

master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?
Code

Files

Permalink
Failed to load latest commit information.
Type
Name
Latest commit message
Commit time
March 29, 2020 16:43
April 30, 2018 04:08
April 11, 2018 11:33
April 30, 2018 04:08
August 5, 2018 14:38
June 28, 2017 09:08
April 30, 2018 04:08
March 15, 2018 17:46
June 24, 2017 15:44
April 30, 2018 04:08
April 30, 2018 04:08
April 22, 2018 21:17
January 6, 2018 23:19
January 10, 2018 15:18
June 4, 2018 08:13
November 28, 2017 21:57
April 20, 2018 14:02
April 14, 2018 23:09
May 1, 2018 15:02

Rasa NLU for Chinese, a fork from RasaHQ/rasa_nlu.

Please refer to newest instructions at official Rasa NLU document

中文Blog

Files you should have:

  • data/total_word_feature_extractor_zh.dat

Trained from Chinese corpus by MITIE wordrep tools (takes 2-3 days for training)

For training, please build the MITIE Wordrep Tool. Note that Chinese corpus should be tokenized first before feeding into the tool for training. Close-domain corpus that best matches user case works best.

A trained model from Chinese Wikipedia Dump and Baidu Baike can be downloaded from 中文Blog.

  • data/examples/rasa/demo-rasa_zh.json

Should add as much examples as possible.

Usage:

  1. Clone this project, and run
python setup.py install
  1. Modify configuration.

    Currently for Chinese we have two pipelines:

    Use MITIE+Jieba (sample_configs/config_jieba_mitie.yml):

language: "zh"

pipeline:
- name: "nlp_mitie"
  model: "data/total_word_feature_extractor_zh.dat"
- name: "tokenizer_jieba"
- name: "ner_mitie"
- name: "ner_synonyms"
- name: "intent_entity_featurizer_regex"
- name: "intent_classifier_mitie"

RECOMMENDED: Use MITIE+Jieba+sklearn (sample_configs/config_jieba_mitie_sklearn.yml):

language: "zh"

pipeline:
- name: "nlp_mitie"
  model: "data/total_word_feature_extractor_zh.dat"
- name: "tokenizer_jieba"
- name: "ner_mitie"
- name: "ner_synonyms"
- name: "intent_entity_featurizer_regex"
- name: "intent_featurizer_mitie"
- name: "intent_classifier_sklearn"
  1. (Optional) Use Jieba User Defined Dictionary or Switch Jieba Default Dictionoary:

    You can put in file path or directory path as the "user_dicts" value. (sample_configs/config_jieba_mitie_sklearn_plus_dict_path.yml)

language: "zh"

pipeline:
- name: "nlp_mitie"
  model: "data/total_word_feature_extractor_zh.dat"
- name: "tokenizer_jieba"
  default_dict: "./default_dict.big"
  user_dicts: "./jieba_userdict"
#  user_dicts: "./jieba_userdict/jieba_userdict.txt"
- name: "ner_mitie"
- name: "ner_synonyms"
- name: "intent_entity_featurizer_regex"
- name: "intent_featurizer_mitie"
- name: "intent_classifier_sklearn"
  1. Train model by running:

    If you specify your project name in configure file, this will save your model at /models/your_project_name.

    Otherwise, your model will be saved at /models/default

python -m rasa_nlu.train -c sample_configs/config_jieba_mitie_sklearn.yml --data data/examples/rasa/demo-rasa_zh.json --path models
  1. Run the rasa_nlu server:
python -m rasa_nlu.server -c sample_configs/config_jieba_mitie_sklearn.yml --path models
  1. Open a new terminal and now you can curl results from the server, for example:
$ curl -XPOST localhost:5000/parse -d '{"q":"我发烧了该吃什么药?", "project": "rasa_nlu_test", "model": "model_20170921-170911"}' | python -mjson.tool
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   652    0   552  100   100    157     28  0:00:03  0:00:03 --:--:--   157
{
    "entities": [
        {
            "end": 3,
            "entity": "disease",
            "extractor": "ner_mitie",
            "start": 1,
            "value": "发烧"
        }
    ],
    "intent": {
        "confidence": 0.5397186422631861,
        "name": "medical"
    },
    "intent_ranking": [
        {
            "confidence": 0.5397186422631861,
            "name": "medical"
        },
        {
            "confidence": 0.16206323981749196,
            "name": "restaurant_search"
        },
        {
            "confidence": 0.1212448457737397,
            "name": "affirm"
        },
        {
            "confidence": 0.10333600028547868,
            "name": "goodbye"
        },
        {
            "confidence": 0.07363727186010374,
            "name": "greet"
        }
    ],
    "text": "我发烧了该吃什么药?"
}

About

Turn Chinese natural language into structured data 中文自然语言理解

Topics

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Sponsor this project

 

Packages

No packages published

Languages