From General to Specialized Domain: Analyzing Three Crucial Problems of Biomedical Entity Disambiguation

Abstract

Entity disambiguation is the task of mapping ambiguous terms in natural-language text to its entities in a knowledge base. Most disambiguation systems focus on general purpose knowledge bases like DBpedia but leave out the question how those results generalize to more specialized domains. This is very important in the context of Linked Open Data, which forms an enormous resource for disambiguation. We implement a ranking-based (Learning To Rank) disambiguation system and provide a systematic evaluation of biomedical entity disambiguation with respect to three crucial and well-known properties of specialized disambiguation systems. These are (i) entity context, i.e. the way entities are described, (ii) user data, i.e. quantity and quality of externally disambiguated entities, and (iii) quantity and heterogeneity of entities to disambiguate, i.e. the number and size of different domains in a knowledge base. Our results show that (i) the choice of entity context that is used to attain the best disambiguation results strongly depends on the amount of available user data, (ii) disambiguation results with large-scale and heterogeneous knowledge bases strongly depend on the entity context, (iii) disambiguation results are robust against a moderate amount of noise in user data and (iv) some results can be significantly improved with a federated disambiguation approach that uses different entity contexts. Our results indicate that disambiguation systems must be carefully adapted when expanding their knowledge bases with special domain entities.

DoSeR - A Knowledge-Base-Agnostic Framework for Entity Disambiguation Using Semantic Embeddings

“…Overall, the results of DoSeR are slightly worse than those of the previous experiment. This is because our index does not only contain named entities and thus, the entity target set Ω comprises more entities to be disambiguated [26]. Using surface forms from the Wikipedia corpus (DoSeR +WikiSF) improves the results from 61.4% F1 to 67.4% F1 on average, which is mainly caused by increased recall values.…”

“…The authors of [26] provide an evaluation of biomedical disambiguation systems with respect to three crucial properties: entity context, i.e. the way entities are described, user data, i.e.…”

“…Thus, these popular entities provide a higher probability to reoccur in other documents [14,26]. The prior p(e i ) describes the a-priori probability that entity e i occurs.…”

Search-based entity disambiguation with document-centric knowledge bases

“…through a description, or extensionally, i.e. through instances and usage [12,23]. Intensional definitions can be understood as a thesaurus or logical representation of an entity, as it is provided by LOD repositories.…”

“…Entity disambiguation with entity-centric KBs from the LOD cloud has been extensively studied on di↵erent domains [16,18,23,24]. In contrast, recent work shows that search-based disambiguation with document-centric KBs attains strong results in the biomedical domain [22,23]). These search-based approaches can be subdivided into two major parts: First, these algorithms retrieve those documents from a document-centric KB, that contain similar textual content as given by the surface form to disambiguate.…”