Individual Submission Summary
Share...

Direct link:

Problems With Sentiment Analysis

Mon, May 29, 11:00 to 12:15, Hilton San Diego Bayfront, Floor: 3, Aqua 309

Abstract

Sentiment analysis is increasingly used across the media and communication field, ostensibly to identify and measure emotional or affective tendencies within user-generated content. We argue that this method conceptually echoes broader shifts toward datafication and ‘post-factual’ politics. We furthermore contend that the expansion of sentiment analysis (and related approaches from the advertising industry) across the media industries has ushered in a new interest in the quantification of feeling, and a new orientation towards affect.

The paper poses three related problems: Sentiment analysis as currently practiced in the media industries has become disconnected from the methodological assumptions and material features of its original design in computational linguistics. This disconnection has produced an application without a concept – measuring ‘sentiment’ frequently fails to establish what sentiment actually signifies. And the adoption of this industry method in academia reiterates the ongoing dynamic tension between administrative and critical research, particularly around inequitable access to data.

Authors