Enhancing text clustering by leveraging Wikipedia semantics

Abstract

Most traditional text clustering methods are based on "bag of words" (BOW) representation based on frequency statistics in a set of documents. BOW, however, ignores the important information on the semantic relationships between key terms. To overcome this problem, several methods have been proposed to enrich text representation with external resource in the past, such as WordNet. However, many of these approaches suffer from some limitations: 1) WordNet has limited coverage and has a lack of effective word-sense disambiguation ability; 2) Most of the text representation enrichment strategies, which append or replace document terms with their hypernym and synonym, are overly simple. In this paper, to overcome these deficiencies, we first propose a way to build a concept thesaurus based on the semantic relations (synonym, hypernym, and associative relation) extracted from Wikipedia. Then, we develop a unified framework to leverage these semantic relations in order to enhance traditional content similarity measure for text clustering. The experimental results on Reuters and OHSUMED datasets show that with the help of Wikipedia thesaurus, the clustering performance of our method is improved as compared to previous methods. In addition, with the optimized weights for hypernym, synonym, and associative concepts that are tuned with the help of a few labeled data users provided, the clustering performance can be further improved.

Reasoning with Heterogeneous Knowledge for Commonsense Machine Comprehension

“…We performed lemmatisation, part of speech annotation, named entity tagging, and dependency parsing using Stanford CoreNLP toolkits . We used the Jan. 30, 2010 English version of Wikipedia and processed it according to the method described by Hu et al (2008).…”

What’s New? Analysing Language-Specific Wikipedia Entity Contexts to Support Entity-Centric News Retrieval

“…Examples include information retrieval [4,14,18], named entity disambiguation [1,2,7,8,11,12], text classification [25] and entity ranking [10]. To extract the content of an entity context, many researches directly used the Wikipedia article describing the entity [1,2,8,9,14,[25][26][27]; some works extended the article with all the other Wikipedia articles linked to the Wikipedia article describing the entity [6,7,12]; while some only considered the first paragraph of the Wikipedia article describing the entity [2]. Different from these approaches, our Graph-based approach not only employs in-links and languagelinks to broaden the article set that is likely to mention the entity, but also performs a finer-grained process: extracting the sentences that mention the entity, such that all the sentences in our context are closely related to the target entity.…”

“…As to the context-based representation vector of the entity, [1,11] defined it as the tf-idf/word count/binary occurrence values of all the vocabulary words in the context content; [2,19] defined it as the word count/binary occurrence values of other entities in the context content; [5,6,9,14,25] defined it as the tf-idf similarity values between the target entity's context content and other entities' context contents from Wikipedia; [27] defined it as the visiting probability from the target entity to other entities from Wikipedia; [7,26] used a measurement based on the common entities linked to the target entity and other entities from Wikipedia. Different from all former researches, we employ aspect weights that have a different interpretation of the frequency and selectivity than the typical tf-idf values and take co-occurrence and language specificity of the aspects into account.…”

“…Some researches [1,2,9,12,14,25] also employed category-links to the Wikipedia article describing the entity. Since the category structure of Wikipedia is language-specific, it is hard to gain insights about cross-lingual context similarity for our case.…”

Clustering Documents Using a Wikipedia-Based Concept Representation

“…Various approaches have been proposed [3,15,5]. We take the same route as [9], and use Wikipedia's vocabulary of anchor texts to connect words and phrases to Wikipedia articles.…”