Innovative approach helps organizations improve knowledge-driven processes, sets a new state of the art in AI
Expert System announced new advancements in applying knowledge graphs and machine learning to natural language processing, consolidating its positioning at the forefront of AI. Published in the Semantic Web journal[1], the results produced by the company’s R&D Lab are endorsed by the largest empirical study around the topic to date and demonstrate how Expert System’s approach outperforms competing efforts, setting a new state of the art in the field.
While machine learning and deep learning have become extremely popular methods, they are still far from addressing natural language understanding, with strong concerns from the AI community in important aspects like explainability and common-sense reasoning. By contrast, because knowledge graphs provide rich, expressive and actionable descriptions of the domain of interest, they offer clear explanations for cognitive processing outcomes. As a result, unstructured data (such as natural language) can be understood and processed faster and more accurately, enabling AI systems to improve knowledge-driven processes through Natural Language Understanding (NLU) and Natural Language Processing (NLP).
“Despite the hype, machine learning alone still fails in natural language understanding in real business scenarios. On the contrary, our knowledge-graph-based approach delivers solid performance in all scenarios” says Marco Varone, Expert System President and CTO. “Our approach is to empower machine learning with the wealth of information embedded in our knowledge graph. This means better results can be reached with less work, and we can approach real natural language understanding.”
Expert System’s market-leading approach is based on scientific research on neural-symbolic AI integration for text and natural language understanding. In particular, the company has been leading the area of knowledge graphs, recently identified by Gartner as one of the emerging technology trends that will blur the lines between human and machine. The breakthrough announced by Expert System now enables machine learning and deep learning models to leverage the same quality of data provided by the Cogito® knowledge graph. This means also that the implementation of AI applications can be optimized as it is possible to combine in an agile way a large and extremely well curated body of structured knowledge with data-driven models extracted from large document corpora.
[1] In print for the next issue of the Semantic Web Journal under title Vecsigrafo: Corpus-based word-concept embeddings – Bridging the statistic-symbolic representational gap in natural language processing. Pre-prints currently available here.
