Chinese English Dictionary of Accounting & Financial sentiment

In the age of big data , More and more financial researchers begin to pay attention to the annual reports of listed companies 、 The tone and mood contained in news media reports and investors' social media posts , And use this to carry out correlation analysis . The premise of such research is to build an emotional Dictionary , It is the basis for measuring and constructing intonation and emotion indicators . The emotional dictionaries used in existing studies generally have the following problems , for example ： Use a general language dictionary instead of a professional financial sentiment Dictionary , This may lead to the omission of key financial sentiment words ; A dictionary of financial sentiment based on small samples is constructed by using manual discrimination method , It may lead to inconsistent judgment standards of emotional words and sample deviation ; Directly use the translated English Dictionary of financial sentiment , This may not capture the different expression habits of different languages for the same emotion ; Use a single class text sample to construct a dictionary , It may result in incompatibility with official financial documents at the same time （ Like news 、 Notice 、 Annual report, etc ） And informal networks （ Ru Gu 、 Forum, etc ） These two types of texts convey the same emotion .

Download link ： Dictionary of accounting and financial sentiment .zip- Dataset document class resources -CSDN download

This paper shares the construction methods and dictionary data of two representative Chinese emotion dictionaries in the financial field , If scholars need to use data, please quote the original text ：

One 、 Yao Wei , Feng Xu , Wang zanjun , Ji Rongrong , Zhang Wei . intonation 、 Emotional and market impact ： Based on the dictionary of financial sentiment . Journal of Management Science ,2021. 24(5), 26-46.

Through text analysis and machine learning, this paper constructs a Chinese emotion dictionary in the financial field . The dictionary construction method has the advantage of avoiding manual judgment as much as possible , From a large sample , And suitable for Chinese text expression and other advantages . The dictionary is aimed at the differences between formal financial texts and social media financial texts , It is divided into formal term emotional dictionary and informal term emotional dictionary . among , The emotional Dictionary of formal terms is suitable for intonation analysis of official texts such as the company's annual report , The informal language emotion dictionary is suitable for emotional analysis of informal texts such as social media .

  Download link ： Dictionary of accounting and financial sentiment .zip- Dataset document class resources -CSDN download

Two 、Bian S , Jia D , Li F , et al. A New Chinese Financial Sentiment Dictionary for Textual Analysis in Accounting and Finance[J]. Social Science Electronic Publishing.

Use HOWNET、DLUTSD、NTUSD Three kinds of dictionaries are used as initial dictionaries , And collected in the line performance summary （online roadshow transcripts）、 Performance statement conference call minutes （earnings conference call transcripts）、IPO The prospectus （IPO prospectus） The corpus of company annual report is constructed . Based on Algorithms and human judgment , Use multistage culling to build “ Chinese Dictionary of financial emotions CFSD”.

Specific steps ：

（1） Merge HOWNET、DLUTSD、NTUSD Three emotional dictionaries , Remove the repetition

（2） Collected 1411 A summary of the performance on the line 、7138 This is a summary of the conference call 、2043IPO The prospectus and 29737 Annual report of the company .jieba Used to split documents , structure “ Basic corpus ”

（3） computational procedure 1 All the words in “ Basic corpus ” The frequency of words in , The frequency of words is 0 The words of , Remove . Words not related to finance are also removed , Finally, it builds “CFSD0.0” Chinese Dictionary of financial emotions .

（4） be-all CFSD0.0 All the words and expressions come from three versions of the general dictionary （HOWNET、DLUTSD、NTUSD）, But these three dictionaries do not contain the positive and negative words that often appear in the financial field . We're going to “CFSD0.0” The most commonly used in the financial field has been added to the emotional dictionary 100 It's a positive word 100 A negative word , build “CFSD0.1” Chinese Dictionary of financial emotions .

（5）Gensim yes python A text analysis library in , This step is mainly used to train word vectors through a large number of corpora . Word vectors can use cosines cos Calculate the similarity . In this step , To calculate the CFSD0.1 The word vector of each word in the version , And then from “ Basic corpus ” Find every word in （CFSD0.1 The words in ） The most similar 50 Word . Get rid of the words that have nothing to do with finance （ Including similar words 、 A synonym for ）, build “CFSD0.2 Chinese Dictionary of financial emotions ”

（6） Merge “CFSD0.0、CFSD0.1、 CFSD0.2”, Remove the repetition , And finally build “CFSD Chinese Dictionary of financial emotions ”

All right CFSD The dictionary has 1489 A negative word ,1108 It's a positive word .