Name Discrimination by Clustering Similar Contexts

Abstract

Abstract. It is relatively common for different people or organizations to share the same name. Given the increasing amount of information available online, this results in the ever growing possibility of finding misleading or incorrect information due to confusion caused by an ambiguous name. This paper presents an unsupervised approach that resolves name ambiguity by clustering the instances of a given name into groups, each of which is associated with a distinct underlying entity. The features we employ to represent the context of an ambiguous name are statistically significant bigrams that occur in the same context as the ambiguous name. From these features we create a co-occurrence matrix where the rows and columns represent the first and second words in bigrams, and the cells contain their log-likelihood scores. Then we represent each of the contexts in which an ambiguous name appears with a second order context vector. This is created by taking the average of the vectors from the co-occurrence matrix associated with the words that make up each context. This creates a high dimensional "instance by word" matrix that is reduced to its most significant dimensions by Singular Value Decomposition (SVD). The different "meanings" of a name are discriminated by clustering these second order context vectors with the method of Repeated Bisections. We evaluate this approach by conflating pairs of names found in a large corpus of text to create ambiguous pseudo-names. We find that our method is significantly more accurate than the majority classifier, and that the best results are obtained by having a small amount of local context to represent the instance, along with a larger amount of context for identifying features, or vice versa.

Resolving Person Names in Web People Search

“…E.g., Pedersen et al [22] propose a method based on clustering using second-order context vectors derived from singular value decomposition (SVD) on a bigramdocument co-occurrence matrix. And Al-Kamha and Embley [1] study combinations of three different representation methods-attribute (factoid) based representations like those used in [20,23], link/citation-based, and content-based.…”

“…One particular case of this people-document association task is referred to as personal name resolution [26,27,23] (also referred to as personal name disambiguation/discrimination [9,22], and cross-document co-reference [4,12]). The task is as follows: given a set of documents all of which refer to a particular person name but not necessarily to a single individual (usually called referent), identify which documents are associated with each referent by that name.…”

Approximate personal name‐matching through finite‐state graphs

“…The disambiguation of authors whose names coincide and overlap in the citation indexing systems calls for methods that distinguish the contexts in which authors or individuals appear, such as cross-document co-reference resolution algorithms that use the vector space model to resolve ambiguities (Bagga & Baldwin, 1998;Gooi & Allan, 2004;Han, Giles, Zha, Li, & Tsioutsiouliklis, 2004;Wacholder, Ravin, & Choi, 1997), co-occurrence analysis and clustering techniques (Han, Zha, & Giles, 2005;Mann & Yarowsky, 2003;Pedersen, Purandare, & Kulkarni, 2005), probabilistic similarity metrics (Torvik, Weeber, Swanson, & Smalheiser, 2005), or co-citation analysis and visualization mapping algorithms through a Kohonen network (X. Lin, White, & Buzydlowski, 2003;McCain, 1990).…”

IdentityRank: Named Entity Disambiguation in the Context of the NEWS Project

“…N ent (11) That is, the average accuracy is defined as the average number of right assignments entity/instance of our naive algorithm divided by the total number of possible assignments. The total number of assignments coincides with the number of entities in the corpus (N ent ) due to the assumption that each entity has at least one candidate.…”

“…Those can be further classified on those that take a "bag of words" context (the position of the words taken as context is not considered) like [11] and those that try to use the role of each word in the context and their relation with the entity [9]. Although some approaches use both common words and named entities as context [11], others suggest that better results can be obtained using as context only other named entities [9]. -The use of knowledge sources like lexical databases, etc., that define the instances that should be matched against the entities and can provide information that can be exploited to perform the matchings.…”