From Text to Knowledge with Graphs: modelling, querying and exploiting textual content

TL;DR

This paper explores how graphs can effectively represent, query, and analyze textual content by integrating linguistics, NLP, and AI to enhance knowledge extraction from diverse text sources.

Contribution

It proposes the hypothesis that annotated graphs are a suitable representation for textual content, combining multiple fields for improved knowledge access and analysis.

Findings

Graphs can effectively model textual content.

Integrating linguistics, NLP, and AI enhances knowledge extraction.

Graphs facilitate advanced querying and analytics of text.

Abstract

This paper highlights the challenges, current trends, and open issues related to the representation, querying and analytics of content extracted from texts. The internet contains vast text-based information on various subjects, including commercial documents, medical records, scientific experiments, engineering tests, and events that impact urban and natural environments. Extracting knowledge from this text involves understanding the nuances of natural language and accurately representing the content without losing information. This allows knowledge to be accessed, inferred, or discovered. To achieve this, combining results from various fields, such as linguistics, natural language processing, knowledge representation, data storage, querying, and analytics, is necessary. The vision in this paper is that graphs can be a well-suited text content representation once annotated and the right querying and analytics techniques are applied. This paper discusses this hypothesis from the perspective of linguistics, natural language processing, graph models and databases and artificial intelligence provided by the panellists of the DOING session in the MADICS Symposium 2022.