NLP_term_EmotionIndex

Term project of NLP. An Emotion Volatility Index uses sentiment analysis(svm) on financial news articles

Training Data:

• in /dataset folder, yelp review data.

• parse.py get all 5-star/1-star review as pos/neg training data

classifier.py

• train a model

• calculate accuracy/precision of the model

News Data

• in /financial-news-dataset-master folder, Bloomberg and Reuters news

• from 2006/10/20 to 2013/11/26

• parse.py

• get every news' date+context

• put them into one file news.json & news.txt

Example

	  '''
		  2010-08-01
			  Bridgepoint Capital Ltd is in advanced talks to buy French 
			  jewelery retailers Histoire dOr and Marc [...] Qualium 
			  Investissement    To contact the reporter on this story  
			  Jonathan Browning  in London   jbrowning9bloombergnet                                         
		  2010-08-01
			  Cell C Pty Ltd South Africas thirdlargest mobile phoneoperator 
			  apologized to customers for poor service and said it was taking 
			  steps to fix the problem    Criticism [...] South Africas 
			  biggest mobilephone operators    To contact the reporter on 
			  this story  Mike Cohen  in Cape Town at   mcohen21bloombergnet
	  '''

news file is to large so splited because it is too large to process

• /xa

• However, there is too many, not able to upload

term.py

• train the model

• use the model to classify all news article

• store to a file of resulted classification

• result.txt

lastParse.py

• process result.txt to assign weights to the positive and the negative each day

• create a file for plot

• final.txt

plot.py

• plot the data

• \plots

NLP term project Emotion_Index