Abstract:
This article is devoted to comparison of various approaches towards sentiment analysis (polarity classification) of the English financial news. The purpose of the work is to check, weather higher quality English texts would have higher classification quality and the links between quality of polarity classification and the architectures of the chosen systems. The approaches, presented on SemEval-2017 competition in the appropriate category of tasks, are taken as the basis. A specially collected dataset is used as the source of English financial news. In the end, a conclusion is made about the relationship between the quality of the source of English financial news and the quality of the news data sentiment analysis, made by the compared systems
Machine summary:
ru2 Kazan Federal University 702 Sentiment Analysis on English Financial News 1271 topic: for example, if we search for the keyword "Windows 8" in Twitter, then we only need to determine the tone of the found tweets.
To gain more insight about sentiments in English financial texts, SemEval, International Workshop on Semantic Evaluation, organized a special task, "FineGrained Sentiment Analysis on Financial Microblogs and News", for the competition of 2017.
The current work was limited to comparison of two competitors, however reevaluating the following systems may be interesting as future work: UW-FinSent at SemEval-2017 Task 5: Sentiment Analysis on Financial News Headlines using Training Dataset Augmentation (Soboleva, 2015) (14 place, 2nd track), link to the implementation: https://github.
com/vln337/semeval2017-task5/blob/master/ Lancaster A at SemEval-2017 Task 5: Evaluation metrics matter: predicting sentiment from financial news headlines (John & Vechtomova, 2017) (4 place, 2nd track), link to the implementation: https :// gi thub.
com/ apmoore I/ semeval HHU at SemEval-2017 Task 5: Fine-Grained Sentiment Analysis on Financial Data using Machine Leaming Methods (Moore & Rayson, 2017) ( 4 place, I st track), link to the implementation: https :// github.
Summary We compared two systems, that were created for solving Task-5 from SemEval-2017, adapting them for new data source with richer English texts, and observed results with higher quality in general and less difference in quality.
HHU at SemEval-2017 Task 5: FineGrained Sentiment Analysis on Financial Data using Machine Learning Methods.
Semeval-2017 task 5: Fine-grained sentiment analysis on financial microblogs and news.
RiTUAL-UH at SemEval-2017 task 5: Sentiment analysis on financial data using neural networks.