BEL: Bagging for Entity Linking

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 2075–2086, Dublin, Ireland, August 23-29 2014. BEL: Bagging for Entity Linking Zhe Zuo, Gjergji Kasneci, Toni Gruetze, Felix Naumann Hasso Plattner Institute Prof.-Dr.-Helmert-Straße 2-3, 14482 Potsdam, Germany {firstname.lastname}@hpi.uni-potsdam.de Abstract With recent advances in the areas of knowledge engineering and information extraction, the task of linking textual mentions of named entities to corresponding ones in a knowledge base has received much attention. The rich, structured information in state-of-the-art knowledge bases can be leveraged to facilitate this task. Although recent approaches achieve satisfactory accuracy results, they typically suffer from at least one of the following issues: (1) the linking quality is highly sensitive to the amount of textual information; typically, long textual fragments are needed to capture the context of a mention, (2) the disambiguation uncertainty is not explicitly addressed and often only implicitly represented by the ranking of entities to which a mention could be linked, (3) complex, joint reasoning negatively affects the efficiency. We propose an entity linking technique that addresses the above issues by (1) operating on a textual range of relevant terms, (2) aggregating decisions from an ensemble of simple classifiers, each of which operates on a randomly sampled subset from the above range, (3) following lo- cal reasoning by exploiting previous decisions whenever possible. In extensive experiments on hand-labeled and benchmark datasets, our approach outperformed state-of-the-art entity linking techniques, both in terms of quality and efficiency. 1 Introduction Named-entity linking (NEL) is the task of establishing a mapping from textual mentions of named entities to canonical representations of those entities in a knowledge base. Often, textual mentions are ambigu- ous; that is, a mention could refer to multiple named entities, but only one of them is correct in the given textual context. Resolving these ambiguities is often referred to as named entity disambiguation (NED), which is a highly challenging aspect of an NEL process. More specifically, a robust NEL algorithm has to robustly resolve ambiguities and thus build on robust NED methods. The NED problem, however, is often ill-posed, as only the right context and background knowledge can help disambiguate entities. In many cases, the contextual information is implicit in nature and may be latently spread across various passages or documents, and background knowledge may not be sufficient, which makes the disambigua- tion task challenging even for human readers. As an example, consider the sentence: “London spent $80,000 ($2,040,000 in current value) to build a 15,000-square-foot stone mansion (‘Wolf House’) on the property.” A human reader knows that in general money is spent by people, but sometimes also city councils can spend money, and hence, in the above sentence “London” may refer to a person or to the city of London. However, when considering the contextual information, especially the key phrase “Wolf House”, and the fact that this was the name of the mansion of the writer Jack London, the disambiguation of “London” becomes obvious. The NED problem is abundant, and the above …

Content truncated to 500 words