How to Make Your Content Smarter
0 likes220 views
This document discusses entity detection and consolidation to enable both human and computer understanding of text. It aims to identify different mentions of real-world entities in text and link them unambiguously to an external database, using linked data as the background database. The researchers developed AELA, a framework for entity detection and consolidation that is adaptive to semantic structures in linked data. Preliminary experiments applying AELA to music and film domains achieved F-scores of 0.54 and 0.87 respectively. Future work will focus on detecting entities mentioned by generalized names, applying AELA to different domains, and evaluating other methods within the AELA framework.
How to Make Your Content Smarter
- 1. Enabling Networked Knowledge Digital Enterprise Research Institute Entity Detection and Consolidation: How to Make Your Content Smarter? Bianca Pereira, Paul Buitelaar Unit for Natural Language Processing Digital Enterprise Research Institute, National University of Ireland, Galway Acknowledgements: This work has been funded by the Science Foundation Ireland under Grant No. SFI/08/CE/I1380 (Lion-2). Motivation: Information available online can be acquired both through human reading and computer processing. Despite this, the majority of data on the Web does not allow both types of reading. Research Questions: • How to identify entity mentions from text? Entity Detection • How to identify which is the real-world entity mentioned on the text? And find the same entity through diverse texts? Entity Consolidation Research Contribution: • Quality assessment of some linked data datasets currently available on the Web. • Identification of common classes and properties used for Named Entities (entities identified by Proper Names) in Linked Data datasets. • Development of a framework adaptable to different linked data datasets. Aim: Link human readable and computer processing content in order to enable machine understanding of the content of a given text and enable humans to track entities across texts. Proposed Solution: • Identification of different mentions to real-world entities in natural language text and their unified, non-ambiguos linking to an external database. • Use the available, and growing, linked data cloud as background database. • Development of AELA, a framework for entity detection and consolidation. Future Research: • Detection of entities mentioned by generalized names (genes, diseases or words such as ambulance, coffee machine, airplane, etc.). • Application of AELA in texts in different domains. • Evaluation of other current methods when applied to AELA. AELA: AELA Framework • Experiments on films and music domains. • Adaptive to the semantic structure of the Linked Data (LD) dataset. Preliminary Results: • Music Domain (Jamendo dataset) F-Score: 0.54 • Films Domain (Linked Movie Database dataset) F-Score: 0.87