Open AccessSentiment Analysis of German Texts in Finance: Improving and Testing the BPW Dictionary
Author(s) -
Matthias Pöferlein
Publication year - 2021
Publication title -
journal of banking and financial economics
Language(s) - English
Resource type - Journals
ISSN - 2353-6845
DOI - 10.7172/2353-6845.jbfe.2021.2.1
Subject(s) - novelty , german , computer science , sentiment analysis , natural language processing , artificial intelligence , word (group theory) , measure (data warehouse) , finance , linguistics , data mining , psychology , economics , social psychology , philosophy
Using the dictionary-based approach to measure the sentiment of finance-related texts is primarily focused on English-speaking content. This is due to the need for domain-specific dictionaries and the primary availability of those in English. Through the contribution of Bannier et al. (2019b), the first finance-related dictionary is available for the German language. Because of the novelty of this dictionary, this paper proposes several reforms and extensions of the original word lists. Additionally, I tested multiple measurements of sentiment. I show that using the edited and extended dictionary to calculate a relative measurement of sentiment, central assumptions regarding textual analysis can be fulfilled and more significant relations between the sentiment of a speech by a CEO at the Annual General Meeting and subsequent abnormal stock returns can be calculated.