Sentiment analysis has recently become core part of analytics in chats between customers and brands specially in ones with measure as customers experience like in hospitality. Lot of research has been done and SaaS solutions are available for finding sentiment of a particular utterance but its hard to understand sentiment for conversations that comprises of conclusive utterances happening between two or more parties. This is an open research area and lots of advancements are needed in finding and mainly classifying interactions into states ( e.g. Greeting, Troubleshooting etc. ). An interesting analytics measure for measuring guests satisfaction will be to find transition of sentiments in utterances as they are happening in conversation.
Conversational sentiments via machine learning and rule based is easier however in goal based chatbots where chatbots are programmed to achieve a particular goal and involve transitioning of conversation from one state to another ( e.g., Greetings -> Problem Discovery -> Troubleshooting -> Conclusion ). Please share any research material you have or advice for finding sentiments for human-human conversations.