Entity-based information retrieval

The goal of this master’s thesis is to extract structured semantic information from Swedish text documents. This information is then intended to be stored as metadata alongside the original text document. Making this metadata searchable should then create a more powerful search engine, both in the sense that it allows for more complex queries and it can be used to give more relevant results. To achieve this, we explored different methods of carrying out named entity recognition. The named entities found in a text can then be used to extract structured information from semantic networks. We associated a unique identifier in a semantic network with each found named entity. The result of this thesis is a program that takes plain text as input and outputs structured information about the entities in the text. We have evaluated the performance of the different parts of the program and compared it to existing systems.