NLP-TFS: Technological Foresight Datamart on Natural Language Processing articles

The NLP-TFS datamart was generated after processing titles and abstracts of 4,714 scientific papers extracted from the Scopus database. From these documents, a total of 110,697 (named) entity occurrences and 11,273 relation occurrences were recognized and categorized by Information Extraction tasks. The dataset image illustrates the Top Named Entity Occurrences panel. The left boxes contain tools to filter the Named Entities (NE) occurrences by country and year and to enable drilling the Entity Dimension to explore each entity category or supercategory separately. From the TF perspective, this functionality allows analysts to look for elements (entities and relations) in texts not by the mere occurrence of text strings but by the role the terms play in the contexts they occur. For example, it is possible to explore elements that are mentioned as materials in the documents. The graph on the right side of the panel shows the NEs ordered by the number of documents in which they occur.