Skip to content

UDICatNCHU/UdicOpenData

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

76 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Open Sentiment Training Data

自己蒐集的training data、字典和stopwords並且包成package,讓大家不用重複造輪子。

Usage

安裝:pip install udicOpenData

  1. 載入實驗室字典, import dictionary
from udicOpenData.dictionary import *
  1. 濾掉stopwords, remove stopwords p.s. rm stop words時就會跟著載入1.實驗室字典了
from udicOpenData.stopwords import *

# default
rmsw(input string, flag=False)

# return segmentation with part of speech.
rmsw(input string, flag=True)
  1. demo:
  • zh:

    >>> doc = '首先,對區塊鏈需要的第一個理解是,它是一種「將資料寫錄的技術」。'
    >>> list(rmsw(doc, flag=True))
    [('區塊鏈', 'n'), ('需要', 'n'), ('第一個', 'm'), ('理解', 'n'), ('一種', 'm'), ('資料', 'n'), ('寫錄', 'v'), ('技術', 'n')]
    
  • en:

    >>> doc = 'The City of New York, often called New York City (NYC) or simply New York, is the most populous city in the United States.'
    >>> list(rmsw_en(doc))
    ['City', 'New York', 'called', 'New York City', 'NYC', 'simply', 'New York', 'populous', 'city', 'United States']
    
    >>> list(rmsw_en(doc, flag=True))
    >>> [('City', 'NNP'), ('New York', 'NNP/NNP'), ('called', 'VBN'), ('New York City', 'NNP/NNP/NNP'), ('NYC', 'NN'), ('simply', 'RB'), ('New York', 'NNP/NNP'), ('populous', 'JJ'), ('city', 'NN'), ('United States', 'NNP/NNS')]
    

For elasticsearch

dump2es.py this command will generate two file with different filename extension

please move these two files into elasticsearch plugin folder

  1. ik:dump2es.py ik
    • stopword:ext_stopword.dic
    • dictionary:mydict.dic
    巨蛋
    遠雄
    趙藤雄
    蔡英文
    陳水扁
    立法院
    蔡正元
    頂新
    食安
    柯p
    ...
    ...
    ...
  2. jieba:dump2es.py jieba
    • stopword:ext_stopword.txt
    • dictionary:mydict.dict
    巨蛋 99
    遠雄 99
    趙藤雄 99
    蔡英文 99
    陳水扁 99
    立法院 99
    蔡正元 99
    頂新 99
    食安 99
    柯p 99
    ...
    ...
    ...

stopword:


上來 
上去 
將不 

www 
http 
https 
.com 




...
...
...

所有語料大小:

  • 正面情緒:約有309163筆,44M
  • 負面情緒:約有320456筆,15M

訓練好的Model:

  1. 政治版Model:
  • 成份: 以下資料皆濾掉標題包含 [公告] 的文章
    • pos.txt 正面情緒的model,將下列版的內容做shuffle,包含:
      • adulation版:標題 + 內文
      • dreams-wish版:標題 + 內文
      • happy版:標題 + 內文
      • kindness版:標題 + 內文裡面的好人行為區段的文字
      • luchky版:標題
    • neg.txt 負面情緒的model,包含:
      • HatePolitics版:標題 + 內文(只有包含黑特且不包含RE的才納入)
  • 大小:
    • pos.txt:13222筆
    • neg.txt:13222筆

產生出資料給 Swinger

python text2json.py positive的檔名(文檔,以一句一句為單位) positive.json(為output檔案的檔名) True/False(若為True就代表要把stopword濾掉 ):會自動把一行一行的語料,斷詞段好給Swinger當input data

About

公開的情緒訓練資料

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages