Kore

Hoffart, Johannes; Seufert, Stephan; Nguyen, Dat Ba; Theobald, Martin; Weikum, Gerhard

doi:10.1145/2396761.2396832

Cited by 150 publications

(38 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We test the embeddings using three EL models on two standard EL datasets, the AIDA-CoNLL 2003 dataset (Hoffart et al, 2012) and the TAC-KBP 2010 dataset (Ji et al, 2010). Two of our EL models are the CNN and RNN EL models used to generate our task-learned embeddings.…”

Section: Methodsmentioning

confidence: 99%

Exploring Neural Entity Representations for Semantic Information

Runge¹,

Hovy²

2020

Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP

View full text Add to dashboard Cite

Neural methods for embedding entities are typically extrinsically evaluated on downstream tasks and, more recently, intrinsically using probing tasks. Downstream task-based comparisons are often difficult to interpret due to differences in task structure, while probing task evaluations often look at only a few attributes and models. We address both of these issues by evaluating a diverse set of eight neural entity embedding methods on a set of simple probing tasks, demonstrating which methods are able to remember words used to describe entities, learn type, relationship and factual information, and identify how frequently an entity is mentioned. We also compare these methods in a unified framework on two entity linking tasks and discuss how they generalize to different model architectures and datasets.

show abstract

Section: Methodsmentioning

confidence: 99%

Exploring Neural Entity Representations for Semantic Information

Runge¹,

Hovy²

2020

Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP

View full text Add to dashboard Cite

show abstract

“…Instead of using individual terms (unigrams), the vector representation may also be made up of n-grams or keyphrases [27,28]. Specifically, Hoffart et al [27] introduce the keyphrase overlap relatedness (KORE) measure, and present approximation techniques, based on min-hash sketches and locality-sensitive hashing, for efficient computation.…”

Section: Term-based Similaritymentioning

confidence: 99%

Semantically Enriched Models for Entity Ranking

Balog

2018

Entity-Oriented Search

View full text Add to dashboard Cite

Most of our efforts in the previous chapter have revolved around constructing termbased representations of entities. These representations can then be ranked using direct adaptations of existing document retrieval models. On the one hand, the resulting approaches are robust and effective across a broad range of application scenarios. On the other hand, these term-based models have little awareness of what it takes to be an entity. Perhaps the most exciting challenge and opportunity in entity retrieval is how to leverage entity-specific properties-attributes, types, and relationships-to improve retrieval performance. This requires a departure from purely term-based approaches toward more semantically informed representations. This change of direction is supported by the emergence of knowledge bases over the past decade (cf. Sect. 2.3). Knowledge bases organize information about entities in a structured and semantically meaningful way. For us, semantics is taken to be synonymous with structure (more precisely, with references to meaningful structure). Our efforts in this chapter are driven by the following question: How can one leverage structured knowledge repositories in entity retrieval?At its core, the entity ranking task (and most IR tasks for that matter) boils down to the problem of matching representations. That is, computing similarities between representations of queries (information needs) and those of entities (information objects). The question then becomes: How to preserve and represent structure associated with entities? Importantly, to be able to make use of richer (i.e., semantic) entity representations during matching, queries also need to have correspondingly enriched representations. For example, if types of entities are represented as semantic units, as opposed to sequences of words, then we also need to know the target types of the query. For now, we shall assume that we are provided with such enriched queries, which will be referred to as keyword++ queries.

show abstract

“…Calculating semantic relatedness between terms is vital in numerous knowledge and information processing tasks of much relevance to the biomedical domain, such as named entity disambiguation (Hoffart et al, 2012), ontology population (Shen et al, 2012), word sense disambiguation (McInnes et al, 2011). Being able to relate entities of interest is crucial in processes of semantic search, information extraction from texts and in building similarity databases.…”

Section: Motivation and Objectivesmentioning

confidence: 99%

EMBnet.journal 19 Suppl. B

journal¹

2013

EMBnet j.

View full text Add to dashboard Cite

Kore

Cited by 150 publications

References 28 publications

Exploring Neural Entity Representations for Semantic Information

Exploring Neural Entity Representations for Semantic Information

Semantically Enriched Models for Entity Ranking

EMBnet.journal 19 Suppl. B

Contact Info

Product

Resources

About