Roughly 80% of data is estimated to be in an unstructured form (Unstructured Big Data). If the goal is to improve know-how with insights from domain experts, AI/NLU algorithms must be adopted as part of the next big data analytics. They can gather contents coming from different type of sources (on-line and off-line), analysing/fusing them by deep semantic NLU algorithms and discovering insights referred to (f.e) particular threats on a dedicated population during a specific temporal reference and in a targeted GEO location. From social media to scientific papers, from technical reports to newspapers, and more; everything becomes a useful source of information so that end-users can find the “information nuggets” that trigger new ideas or activate verifications or interventions. The ability to automatically understand natural language enables new approaches to advanced analysis; even citizens can become “sensors” for their territory by means of analysis of the human factors using avant-garde behavioural algorithms such as emotions and stylometric analysis. The target is: •to automatically read and understand text from acquired documents (es: financial statements and related notes, web sites, news, internal docs…); •to automatically analyse and classify all types of documents and add semantic tags for searching (es: initiative for sustainability, Sentiment, Stakeholder, Citizenship, Governance…); •to extract and normalize key information for ESG domain (es: Certifications, Emissions, Investments, GRI indicators…); •to calculate the Reputation index combining emotional and behavioural scores and the level of severity of possible crimes; •to finally provide a platform for semantic searching and dashboarding so to support the analysis of reputation information and reporting for performance information.
Understanding these indicators is relevant in the selection of partners, customers or suppliers. They are measured in climate change, corporate governance and human capital (ESG factors). The reputation index of a company is crucial in Risk Management/Assessment processes.
Elementi di innovazione
An Hybrid NLU engine based either on Ontology/Taxonomy or ML/DL algorithms.
A tuned Ontology on a specific domain f.e. the Financial Industry Business Ontology (FIBO).
Behavioural algorithms so to extract opinions such as f.e Trust, Fear, Confusion, condemnation for environmental disaster ...
Impatti / Risultati attesi
The approach of extracting relevant semantic tags can greatly increase the “Intelligence” as “Information Superiority” so going to unveil the “exceptions of the rules”, low signals/low level indicators and finally improve and fastening the quality of decision-making processes (Augmented DSS).
Expert.ai is working on ESG with different customers (high TRL).
The Hybrid approach:
•is based on Explainable AI by design because it can be configured, understood and totally managed by semantic rules.
•consumes less energy because it needs less training data, so in line with green technology AI.
Elementi di replicabilità
•recommending truly relevant scientific articles to researchers
•combating identity theft by analysis on line contents and historical use cases
•tracking of illicit money flows by either Surface or Deep/Dark web analysis
•waste monitoring by analysing citizen opinions on social networks
Stato di implementazione
Expart.ai provides high level TRL NLU engines with strong tuning on the specific domains and languages, so that the analytical task can be domain focused, can run automatically and with objectivity on 24/7, going to leave the final deduction to humans.
No partners are involved in the providing of the NLU engine.That’s true Expert.ai always works with partners/customers so to insert it into existing solutions so to protect their investment in the case.