Named Entity Disambiguation using Linked Data

Identification of Named Entities (NE) such as people, organisations and locations is fundamental to semantic annotation and is the starting point of more advanced text mining algorithms. For instance, sentiment analysis is widely used in finance to extract the latest signals and events from news that could affect stock prices. However, before extracting company-related sentiment, it is necessary to identify the documents containing the corresponding and unambiguous company entities. Humans usually resolve ambiguities based on context. We argue that Linked Data can be a valuable source for extending the already available context. We combine a state-of-the-art named entity tool with novel Linked Data-based similarity measures and show that our algorithm can improve disambiguation accuracy on a subset of Wikipedia user profiles.