We connect sociological theory with computational methods to analyse hostile narratives

“

Hate speech detection is a growing industry that attracts significant investment. Moreover, hate speech moderation is important for maintaining the value of social media companies. But most fundementatlly, the Web and society we want should be free from the fear and consequences of hate speech. Current sytems for detecting hate, and indeed other NLP technologies are revolutionising how we interact with machines., but are do they work well for analysing texts for meaning. Personal assistants can convert natural language inputs to generally useful outputs with reasonable reliability. But the same technologies powering these revolutionary technologies are also found in sytems used to extract meaning from text. The meaning extraction systems do not appear to be connected to sociological theories - consider whether there is actually any underpinning theory of sentiment which informs the design of sentiment analysis systems. Our reserach seeks to improve NLP by developing systems that are informed by theories from Peace Research. To do so, we believe, means creating systems that have improved explanatory value than the current state-of-the-art. We would hope this system will improve how hate speech is controlled online in order to get the Web we want.

”