Title:

Correlating Sentiment with Journal Article Success

Journal articles traditionally use a neutral tone to maintain objectivity and avoid bias. Despite this assumption, we hypothesize that subtle differences in voice, diction, and rhetoric can influence the success of journal articles since effective communication is necessary to make a lasting impact. This project tests that hypothesis by correlating sentiment to journal article success. Sentiment analysis from natural language processing is a quantitative proxy for complex and nuanced linguistic qualities. Total citation count is the success metric, and papers are compared with others in their field. Articles were selected from Computer Science, Political Science, and Art to provide a range of quantitative and qualitative research. For example, in Computer Science, quantitative results speak for themselves, while qualitative analysis and persuasive argumentation are necessary for Art. The Semantic Scholar Open Research Corpus on GitHub is the paper database, and Amazon Comprehend is the sentiment analysis model. Statistical analysis of the data showed that while non-neutral sentiment is more variable in Art and Political Science than Computer Science, no successful papers in any field had a substantially non-neutral sentiment score. However, manual investigation shows that the model does not provide enough granularity to show correlation. Further research with tight parameters around the field of study and audience is necessary to test the validity of our approach. Examples with interesting abstracts accompanied by their sentiment score can be found here: https://docs.google.com/document/d/193G5C2821Enz9XCc8PNhm9ffDMxReeYtXWX127j2cLA/edit?usp=sharing We hope it sparks curiosity so you can explore this topic on your own!

Name

Comment*

Comments are viewable only by submitter

Captcha*