Big Data Days 2022

21-24 ноября

Москва и онлайн

Vsevolod Dyomkin

Research Engineer

Franz, Ukraine

Биография

Vsevolod Dyomkin is a Lisp programmer and researcher in the area of Natural Language Processing. He works as a consultant for Franz Inc. on the development of AllegroGraph RDF triple-store and various custom projects involving ML and NLP. Prior to Franz, he was a tech lead of Grammarly’s core NLP team and a technology entrepreneur. Vsevolod is the author of the Apress book «Programming Algorithms in Lisp» and a developer of several training courses on Natural Language Processing, Algorithms, and System Programming. He is a long-time open-source contributor and a co-founder of lang-uk project that aims to collect open datasets for Ukrainian NLP. http://vseloved.github.io/

Доклад

Text2graph — A Practical Approach to Transforming Free-Form Text Into a Computable Graph Structure

Despite all the recent progress in Natural Language Processing associated with the advent of big Deep Learning models such as BERT and GPT, the problem of extracting computable meaning from free-form text remains unsolved (and also rather under-researched). Over the years, a number of technologies have been proposed to predictably turn text into a graph structure, such as Semantic Parsing, AMR, and others. Yet, none of them managed to gain traction and adoption outside niche tasks. In this talk, we describe a novel hybrid ML/symbolic approach to transforming natural language sentences into RDF graph structures, as well as possible practical applications of the results in downstream tasks, such as question answering, text entailment, and others.

Ключевые слова

Natural Language Processing
Graph Databases

« Hазад