RECOGNIZING TRUST IN NATURAL LANGUAGE

Trust in the speaker’s character, his credibility or ethos, is one of the most important and influential elements of real-life practice – from everyday communication, social debates, news, reviews for online sellers or travel sites, legal discourse to human intelligence. Being able to properly evaluate an information source is crucial for deciding whether or not we should believe news report provided by a journalist or a blogger, or whether or not we should convict a suspect on a
basis of a given witness testimony. The goal of the project is to understand how people express trust in natural language and to build a prototype of a system for ethos mining, i.e. semi-automatic and automatic processing and exploring through big maps of arguments linked to the credibility of their providers.

Processing and evaluation of trust in speaker’s credibility is crucial for many different sectors of industry. For example, reviews provide the main source of information for users of online shopping networks such as Amazon and eBay, and travel sites as TripAdvisor. Yet analysing and structuring vast amounts of specialized data can be a very expensive task in terms of time and effort. The main technological challenge is to enable processing big datasets in a more structured and (semi-)automatic way. The area of Computational Ethos is highly innovative and requires early-stage, higher-risk technology development that can lead to further commercialisation in the not-too-distant future.