Arsemotica: what is this

Aims, desired output, approach

aims

Context: Affective computing + Social Semantic Web framework

Aim: extracting a rich emotional semantics (i.e. not limited to a positive or a negative reception) of tagged resources. Sentiment Analysis and beyond!

Desired output: not merely a general appreciation but a set of specific concepts, that emerge as the most significant in capturing the emotions of the users towards the resource at issue, each enriched with a score expressing the strength of that emotion.

Approach: ontology-driven sentiment analysis. Given a tagged resource, the correlation between tags and emotions is computed by exploiting and combining Semantic Web tools and lexical resources.The extraction of the emotional content is driven by an ontology of emotional categories: from unstructured texts to concepts of an ontology. Intuitively:

  • tags directly referring to ontological concepts are identified;
  • potentially affective tags can be identified by users by using the ontology;
  • the final emotional output for the given resource is calculated based on identified emotional concepts by exploiting automated reasoning on ontology relationships.

The whole process is organized in 4 main phases which are sketched in the architecture commented below.

the architecture

  • Phase I. Pre-processing: Lemmatization and String sanitizing
  • Phase II. Checking tags against the ontology of emotions:
    If yes, the tag is “emotional” (directly referring to an emotional category). The information is stored in a DB table as a set of triples (t, e, s), meaning that tag t is related to emotion e with a strength value s ∈ [0, 100] (e.g ‘affanno’, ‘Ansia’, 100).
  • Phase III. Sentiment analysis and user’s feedback:
    Tags not belonging to the ontology are further analysed by means of SentiWordNet. When at least one meaning of the word has a relevant sentimental score (positive or negative), it is offered to the user for a feedback. Users can indicate one or more emotions from the ontology with a strength value (the user’s measure for the semantic affinity of the term with the chosen emotion). Again, a set of triples (t, e, s) is collected.
  • Phase IV. Computing the prevalent emotions:
    Once all the triples are collected, ArsEmotica combine and rank the resulting emotions by reasoning on the ontology hierarchy.

contributors

The core idea of the Arsemotica project was a brainchild of researchers from the Dipartimento di Informatica, and creative members of the Associazione Culturale Arsmeteo (ACA).

The current version of the software Arsemotica (prototype v1.0) has been developed by researchers and students from Dipartimento di Informatica, University of Turin.