Description

TrustServista is an Advanced Digital News Analytics & Verification Platform that performs a series of automated tasks based on web digital content (URLs), which would identify the origin (“Patient Zero”) of a specific news topic, the links between various articles on the same topic, the similarities between them and would generate an automatic trustworthiness/quality score of each content piece using Intelligent Algorithms.

The platform can be deployed on the cloud or on-premises, can process huge amounts of data daily, and supports a wide range of languages.

The processes that are automated by TrustServista closely follow the manual processes of investigative journalism and critical reading. Finding “Patient Zero” is one of the most important indicators of the trustworthiness of a news topic; it indicates where the information originated, which are the sources, and also constitutes a basis for following the propagation of the information, alterations in content, and the effect it has on mainstream or social media.

Embedding more than a dozen NLP algorithms (IntelliDockers), TrustServista has been proven to effectively flag untrusted/low-quality content in real-time, helping analysts shorten investigation times with a 3-step verification approach:

• Publisher Veracity: Each analyzed webpage contains valuable meta-data that impacts the veracity of the publisher: a named author, a verifiable digital “fingerprint” (public DNS ownership, unique advertising code snippets), and contact information. In time, with enough content collected from publishers, a statistical publisher “persona” can be defined, answering the question of “Who is publishing this?”
• Information Source: Every article published online has one or multiple sources of information: news agency wires, other websites, social media, or journalist-recorded events or statements. Automatically building a relationship graph from each article – containing the links to other webpages, references, and close similarities with other articles – can be done using a combination of NLP, Graph Theory, and Semantic Comparison. This way, all the information sources of an article can be traced back to “Patient Zero”, the potential origin of information answering the complicated question of “Where does this come from?”
• Content Quality: The 3rd and most important verification is done on the content itself, as the writing style is directly correlated with trustworthiness. Making heavy use of NLP algorithms, the polarity of the article can be determined (positive, negative, neutral), measuring the author’s subjectivity; the amount of factual information can be extracted as a context-setting (Who? When? Where?) formed by named entities. A unique statistical algorithm can determine if the article is written in a click-baiting style and the content is automatically categorized according to the IPTC and IAB taxonomies. This way, the content analysis can reveal the answer to the question “How was it written?”

An overall trustworthiness score can be calculated based on the Publisher Veracity, Information Source, and Content Quality metrics, automatically and with no human assistance, relying entirely on AI software algorithms.