Scholarly record

SENTIMENT AND TOPICAL ANALYSIS FOR FRENCH TWITTER: THE CASE OF INTER-ETHNIC CONFLICTS

S. S. Bodrunova, I. S. Blekanov, Y. Danilova, N. Zhuravleva, M. Kukarkin

Abstract

Sentiment analysis for the French language remains an under-developed research area. At the same time, sentiment analysis works well for identifying conflictual discourses in English. Today, identifying sentiment in social networks and microblogs, including Twitter, is a growing area of studies, but despite French is one of the most popular post-colonial languages online, the studies of user sentiment in this language are still rare enough. Also, there are practically no studies that would link sentiment analysis in French to other methods of textual exploration, while such methods would be telling of how the sentiment distributes depending on topicality or other text features in large text collections. To assess how sentiment distributes among an emotional discussion, we use the case of Charlie Hebdo massacre of 2015. We employ web crawling, sentiment analysis with automated lexicon, human coding, and machine learning, as well as two approaches to topic modeling in order to see how to detect sentiment distribution in francophone discussion. With the data collected by #jesuischarie, we show that sentiment clusters the discussions and may serve as a ground for divergence of topicality.

Publication details

Back to publication list