Tweet reaction action

Tweets provide a good view on what is being spoken about, but are often too short to allow AI to correctly analyse the intended sentiment. By considering tweets within the context of a conversation, we hope to allow AI to see which way the wind is blowing.

Goal

In this project, we will analyse tweets automatically in order to gain insight in the emotions present in Twitter-conversations, and when these are part of deliberate campaign to incite. This would allow earlier cognisance of attempted instigation.

Results

- Analysis of syntactic and semantic possibilities to extract sentiment from utterances within online conversations
- A prototype sentiment-analyser able to map out the flow of sentiment within a Twitter conversation
- Conceptualisation of indicators for incitement expressed in formal logic
- A short paper motivating and elaborating the possibilities and applications of the prototype
- Evaluation and analysis of results

 

 

Duration

01 June 2021 - 30 June 2022

Approach

Tweets are analysed using accepted NLP methods, and predictions are made on how further conversation would progress. This prediction is verified against actual replies, zooming in on discrepancies.

HU researchers involved in the research

Related research groups

Any questions or want to collaborate?

Related