Product reviews which are increasingly commonplace on the web typically contain a textual component and a numerical rating. The textual component can be viewed as a collection of arguments for and against the product. Whilst the reviewer may not have provided the attacks between these arguments they typically provide an indication of which set of arguments they view as being more acceptable/winning via the numerical rating (i.e. a positive rating indicates that the positive arguments are accepted and vice versa). Our framework builds upon this intuition and we propose a two step process for identifying a probability distribution over the set of possible argument graphs that the reviewer may have had in mind. The first is the identification step in which for a given review, we identify a distribution by analysing the relationship between the rating and polarity of arguments in the review via the constellations approach to probabilistic argumentation. The second step is the refinement step in which we harness ratings from multiple reviews and use this to refine our probability distribution thus enabling us to learn from the data. We illustrate the applicability of our approach by testing it with real data.

In the real world it is common for agents to posit arguments concerning an issue but not directly specify the attack relations between them. Nonetheless the agent may have these attacks in mind and instead they may provide a proxy indicator through which one can infer the agent's intended argument graph (arguments and attacks). Consider online reviews, where reviews are collections of arguments for and against the product (positive and negative) under review and the rating indicates whether the positive or negative arguments succeed ultimately. In previous work [1] we have proposed a method that formalises this intuition and uses the constellations approach to probabilistic argumentation to construct a probability distribution over the set of arguments graphs the agent may have had in mind. In this paper we extend this proposal and provide a method, that uses Bayesian inference, to update the initial probability distribution using real data. We evaluate our proposal by conducting a number of simulations using synthetic data.