Five main stages have been identified for SEEMPAD:
- Create and experiment with debates, test efficiency of the debatable topics
- Design and setup the experimental environment (forum, sensors, etc.)
- Propose a semantic web model to integrate and synchronize all the data.
- Carry out experiments, record and synchronize all data
- Publish the data set on the web using the semantic web formalisms and linked open data practices.
- Process in an automated way the dataset with Natural Language Processing tools, apply argumentation theory to extract the overall debates structures and study the relations with the automatically detected emotions.
For the first year we intend to setup and test a first debatable situation as a proof of concept. The goal is to model a first debate scenario, and to experiment it with a group of participants. Different approaches will be applied in order to investigate suitable techniques to analyze each dimension. In particular, we intend to go through the following steps:
- Create the debate scenario (topic, experimental material, questions, users’ profiles, etc.) and test an experimental protocol for the debate scenario
- Run the experiment and collect data from all the dimensions (i,e. textual, dialogical and emotional)
- Synchronize and annotate the datasets with respect to each dimension
- Textual analysis: recognize the polarity of the arguments using sentiment analysis techniques (relying for instance on lexical resources for opinion mining as SentiWordNet), and integrate such information when detecting the relations among the arguments.
- Argumentation analysis: Construct the overall structure of the debate highlighting the polarity of the relations among the different arguments, and enrich existing argumentation semantics with the emotional aspects underlying the debates.
- Emotional analysis: identify and measure the emotional states occurring during the dialogue; link the emotions to the referring arguments, argumentation steps, and arguments; show the influence of emotional moderator (agent) which can intervene to reduce the hostility and negative emotions.
- Evaluate feasibility of synchronizing the results of the different analysis