An approach to web-scale named-entity disambiguation

2011

Disambiguating named entities in naturallanguage text maps mentions of ambiguous names onto canonical entities like people or places, registered in a knowledge base such as DBpedia or YAGO. This paper presents a robust method for collective disambiguation, by harnessing context from knowledge bases and using a new form of coherence graph. It unifies prior approaches into a comprehensive framework that combines three measures: the prior probability of an entity being mentioned, the similarity between the contexts of a mention and a candidate entity, as well as the coherence among candidate entities for all mentions together. The method builds a weighted graph of mentions and candidate entities, and computes a dense subgraph that approximates the best joint mention-entity mapping. Experiments show that the new method significantly outperforms prior methods in terms of accuracy, with robust behavior across a variety of inputs.

2007

Named entity disambiguation has been one of the main challenges to research in Information Extraction and development of Semantic Web. Therefore, it has attracted much research effort, with various methods introduced for different domains, scopes, and purposes. In this paper, we propose a new approach that is not limited to some entity classes and does not require wellstructured texts. The novelty is that it exploits relations between co-occurring entities in a text as defined in a knowledge base for disambiguation. Combined with class weighting and coreference resolution, our knowledge-based method outperforms KIM system in this problem. Implemented algorithms and conducted experiments for the method are presented and discussed.

Proceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining, 2013

Named entity disambiguation is the task of disambiguating named entity mentions in natural language text and link them to their corresponding entries in a knowledge base such as Wikipedia. Such disambiguation can help enhance readability and add semantics to plain text. It is also a central step in constructing high-quality information network or knowledge graph from unstructured text. Previous research has tackled this problem by making use of various textual and structural features from a knowledge base. Most of the proposed algorithms assume that a knowledge base can provide enough explicit and useful information to help disambiguate a mention to the right entity. However, the existing knowledge bases are rarely complete (likely will never be), thus leading to poor performance on short queries with not well-known contexts. In such cases, we need to collect additional evidences scattered in internal and external corpus to augment the knowledge bases and enhance their disambiguation power. In this work, we propose a generative model and an incremental algorithm to automatically mine useful evidences across documents. With a specific modeling of "background topic" and "unknown entities", our model is able to harvest useful evidences out of noisy information. Experimental results show that our proposed method outperforms the state-of-the-art approaches significantly: boosting the disambiguation accuracy from 43% (baseline) to 86% on short queries derived from tweets.

2015

When RDF instances represent the same entity they are said to corefer. For example, two nodes from different RDF graphs 1 both refer to same individual, musical artist James Brown. Disambiguating entities is essential for knowledge base population and other tasks that result in integration or linking of data. Often however, entity instance data originates from different sources and can be represented using differ- ent schemas or ontologies. In the age of Big Data, data can have other characteristics such originating from sources which are schema-less or without ontological structure. Our work involves researching new ways to process this type of data in order to perform entity disambiguation. Our approach uses multi-level clustering and includes fine-grained entity type recognition, contextualization of entities, online processing of which can be supported by a parallel architecture.

Proceedings of the 21st Conference on Computational Natural Language Learning (CoNLL 2017)

We address the task of Named Entity Disambiguation (NED) for noisy text. We present WikilinksNED, a large-scale NED dataset of text fragments from the web, which is significantly noisier and more challenging than existing newsbased datasets. To capture the limited and noisy local context surrounding each mention, we design a neural model and train it with a novel method for sampling informative negative examples. We also describe a new way of initializing word and entity embeddings that significantly improves performance. Our model significantly outperforms existing state-ofthe-art methods on WikilinksNED while achieving comparable performance on a smaller newswire dataset.

Detecting entity mentions in a text and then mapping them to their right entities in a given knowledge source is significant to realization of the semantic web, as well as advanced development of natural language processing applications. The knowledge sources used are often close ontologiesbuilt by small groups of experts -and Wikipedia. To date, state-of-the-art methods proposed for named entity disambiguation mainly use Wikipedia as such a knowledge source. This paper proposes a method that enriches a close ontology by Wikipedia and then disambiguates named entities in a text based on that enriched one. The method disambiguates named entities in a text iteratively and incrementally, including several iterative steps. Those named entities that are identified in each iterative step will be used to disambiguate the remaining ones in the next iterative steps. The experiment results show that enrichment of a close ontology noticeably improves disambiguation performance.

Abstract. The rapidly increasing use of large-scale data on the Web has made named entity disambiguation a key research challenge in Information Extraction (IE) and development of the Semantic Web. In this paper we propose a novel disambiguation framework that utilizes background semantic information, typically in the form of Linked Data, to accurately determine the intended meaning of detected semantic entity references within texts.

The Semantic Web-ISWC 2006, 2006

Precisely identifying entities in web documents is essential for document indexing, web search and data integration. Entity disambiguation is the challenge of determining the correct entity out of various candidate entities. Our novel method utilizes background knowledge in the form of a populated ontology. Additionally, it does not rely on the existence of any structure in a document or the appearance of data items that can provide strong evidence, such as e-mail addresses, for disambiguating authors for example. Originality of our method is demonstrated in the way it uses different relationships in a document as well as in the ontology to provide clues in determining the correct entity. We demonstrate the applicability of our method by disambiguating authors in a collection of DBWorld posts using a large scale, real-world ontology extracted from the DBLP. The precision and recall measurements provide encouraging results.

2018

Named entity disambiguation (NED) is a central problem in information extraction. The goal is to link entities in a knowledge graph (KG) to their mention spans in unstructured text. Each distinct mention span (like John Smith, Jordan or Apache) represents a multi-class classification task. NED can therefore be modeled as a multitask problem with tens of millions of tasks for realistic KGs. We initiate an investigation into neural representations, network architectures, and training protocols for multitask NED. Specifically, we propose a task-sensitive representation learning framework that learns mention dependent representations, followed by a common classifier. Parameter learning in our framework can be decomposed into solving multiple smaller problems involving overlapping groups of tasks. We prove bounds for excess risk, which provide additional insight into the problem of multi-task representation learning. While remaining practical in terms of training memory and time requirem...

En este trabajo presentamos un sistema no supervisado para agrupar los resultados proporcionados por un motor de búsqueda cuando la consulta corresponde a un nombre de persona compartido por diferentes individuos. Las páginas web se representan mediante n-gramas de diferente información y tamaño. Además, proponemos un algoritmo de clustering capaz de calcular el número de clusters y devolver grupos de páginas web correspondientes a cada uno de los individuos, sin necesidad de entrenamiento ni umbrales predefinidos, como hacen los mejores sistemas del estado del arte en esta tarea. Hemos evaluado nuestra propuesta con tres colecciones de evaluación propuestas en diferentes campañas de evaluación para la tarea de Desambiguación de Personas en la Web. Los resultados obtenidos son competitivos y comparables a aquellos obtenidos por los mejores sistemas del estado del arte que utilizan algún tipo de supervisión. Palabras clave: aprendizaje no supervisado, clustering, n-gramas, búsqueda de personas en la web