We have worked on the following papers relating to emotion and sentiment analysis:
We performed comprehensive analysis on the use of embedded word vectors in sentiment analysis. We have submitted our paper to a conference and are waiting to see if it will be accepted. Then we will publish it here.
One of our conclusions was the need for an organisation to maintain high quality, regularly updated vector spaces. This is because the main vector spaces currently used are quite old: Stanford's Common Crawl GloVe vectors from 2014 and Google News vectors from 2013. We would like to obtain funding to do this.
Before starting this company, we contributed to a research paper from Stanford University titled "Conversational Agents and Mental Health: Theory-Informed Assessment of Language and Affect". The paper is available for download from Stanford's archive. It was submitted and accepted for presentation at the fourth international conference on Human Agent Interaction in Singapore in early October 2016.
The paper analysed how humans feel when they chat with other humans and with artificial bots. They found that humans mirror positive sentiments much more than negative ones when chatting: 84% vs 22% when chatting to other humans, 75% vs 41% when chatting to bots. Furthermore, humans are much more likely to be negative with bots than humans. Our involvement in this study was to automatically extract emotions from conversations.