Discovering relationships among categories using misclassification information

Mengle, Saket S. R.; Goharian, Nazli; Platt, Alana

doi:10.1145/1363686.1363899

Cited by 10 publications

(6 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One should note that much more relationships were identified between categories in the ODP46 dataset than in the 20NG dataset. This finding was reported in [10] and was also validated by human evaluators. As the precision for predicting the relationships in the 20NG dataset (59.0%) is higher than in the ODP46 dataset (31.8%), the probability of wrongly lowering a true positive is higher in the OPD46 dataset.…”

Section: Effects Of Misclassification Informationsupporting

confidence: 80%

“…While performing text classification on the documents, we keep track of the number of keywords [10] in each document. Keywords are terms that are found to consistently occur in documents of a category, and are not frequently found in documents of other categories.…”

Section: Least Ambiguous Segmentsmentioning

confidence: 99%

“…However, some datasets do not have a defined ontology. Hence, we explore an automatic method to discover relationships among categories [10]. The misclassification information generated during the process of text classification is used to predict relationships among categories.…”

Section: Misclassification Informationmentioning

confidence: 99%

See 2 more Smart Citations

Improving classification based off-topic search detection via category relationships

Platt

Mengle

Goharian

2009

Proceedings of the 2009 ACM Symposium on Applied Computing

Self Cite

View full text Add to dashboard Cite

The illegitimate access of documents by insiders (also known as off-topic search) is an increasingly prevalent and largely ignored problem. We propose an approach that uses text classification for off-topic search detection. Our empirical results indicate that offtopic search detection effectiveness improves by considering only a subset of documents that are retrieved for a given user query. Furthermore, we also show that the effectiveness of off-topic search detection improves by using the ontological information of document categories. Our empirical results demonstrate that utilizing sibling relationship information and relationships derived from misclassification information statistically significantly improves the results over the baseline in most cases.

show abstract

Section: Effects Of Misclassification Informationsupporting

confidence: 80%

Section: Least Ambiguous Segmentsmentioning

confidence: 99%

See 1 more Smart Citation

Improving classification based off-topic search detection via category relationships

Platt

Mengle

Goharian

2009

Proceedings of the 2009 ACM Symposium on Applied Computing

Self Cite

View full text Add to dashboard Cite

show abstract

“…Due to the large amount of online news documents, effective management of these resources has become a challenging goal for researchers in the field of information retrieval, text categorization and text mining. Towards this, the most common functions include clustering and classifying news documents [4,7,26,35,39], extracting semantic relationships between entities from news documents [37,40,41], detecting event from news stories [3,18,28,34], and discovering meaningful relations among news documents [20,30,38]. Among these tasks, discovering news relations has recently been focused among several studies.…”

Section: Introductionmentioning

confidence: 99%

Region-based association measures for ranking mined news relations

Kittiphattanabawon

Theeramunkong

Nantajeewarawat

2014

IDA

View full text Add to dashboard Cite

Association rule mining can be applied to discover relations among news documents. Most existing approaches may not be good enough to extract meaningful news relations due to the limitation of having only single association measure for ranking mined news relations. This paper presents a region-based method to selectively use different association measures for different ranking regions, towards improvement of the ranking mechanism for news relation discovery. In this method, first the mined relations are sorted under a preliminary criterion to form a number of regions before scoring the relations in each region with different association measures. The meaningful news relations are discovered through three levels of relation: completely related, somehow related and unrelated relations, judged by the domain expert. To evaluate the proposed regionbased ranking method, the method which has no region construction is considered to be the baseline by using a set of 1,132 news relations mined from 811 news documents. As performance criterion, a rank-order mismatch is explored to compare the qualitative results between the proposed method and the human evaluation. Compared to the baseline, the region-based method significantly improves performance by the average rank-order mismatch of 1.21%-28.32% for confidence and 4.83%-29.04% for conviction, respectively. N. Kittiphattanabawon et al. / Region-based association measures for ranking mined news relationsresulting in the complexity of O(N L ), where L is the length of the longest pattern. However, an inevitable issue of association rule mining is triggered by the large number of rules generated [5,22,23,27,31]. In the past, several works introduced various approaches to solve the problem of these tremendous association rules, which may be classified into filtering, constraint-based mining and ranking.Up to present, most previous works have applied single objective measure (such as support, confidence or lift) or single subjective measure (such as preference, personal value or cultural value) or their combinations, to filter, constrain, or rank the discovered rules. However, they still focus on a uniform function for the whole ranking interval. Intuitively, only a single objective function for the whole range of candidates may not be enough to prioritize the rules. It is worth exploring multiple objective functions for rule prioritization. Towards such issues, this paper presents a method to selectively use different objective functions (association measures) for different ranking regions to improve the ranking mechanism. So-called region-based ranking, the proposed approach, sets a specific weight for each different region (range) of the relation strength in order to maximize classification performance. This work also deals with two main issues, i.e., (1) how to determine suitable points of weight scheme change in order to form regions, and (2) which weight scheme we should assign to each region. In addition, when the domain is changed to other areas such as medical, enviro...

show abstract

“…Towards automated content organization, while classification techniques can be applied to assign a category label to each document based on a number of criteria, such as text genre, text style, and users' interest [1]- [3]. Some of them can be adopted for classifying news documents [4], [5]. By the classification method, it requires users to provide a number of predefined classes and a large number of training examples.…”

Section: Introductionmentioning

confidence: 99%

News Relation Discovery Based on Association Rule Mining with Combining Factors

Kittiphattanabawon

Theeramunkong

Nantajeewarawat

2011

IEICE Trans. Inf. & Syst.

View full text Add to dashboard Cite

Nichnan KITTIPHATTANABAWON†a) , Student Member, Thanaruk THEERAMUNKONG †b) , and Ekawit NANTAJEEWARAWAT †c) , Members SUMMARY Recently, to track and relate news documents from several sources, association rule mining has been applied due to its performance and scalability. This paper presents an empirical investigation on how term representation basis, term weighting, and association measure affects the quality of relations discovered among news documents. Twenty four combinations initiated by two term representation bases, four term weightings, and three association measures are explored with their results compared to human judgment of three-level relations: completely related, somehow related, and unrelated relations. The performance evaluation is conducted by comparing the top-k results of each combination to those of the others using so-called rank-order mismatch (ROM). The experimental results indicate that a combination of bigram (BG), term frequency with inverse document frequency (TFIDF) and confidence (CONF), as well as a combination of BG, TFIDF and conviction (CONV), achieves the best performance to find the related documents by placing them in upper ranks with 0.41% ROM on top-50 mined relations. However, a combination of unigram (UG), TFIDF and lift (LIFT) performs the best by locating irrelevant relations in lower ranks (top-1100) with 9.63% ROM. A detailed analysis on the number of the three-level relations with regard to their rankings is also performed in order to examine the characteristic of the resultant relations. Finally, a discussion and an error analysis are given. key words: news relations, news relation discovery, association rule mining, combining factors

show abstract

Discovering relationships among categories using misclassification information

Cited by 10 publications

References 8 publications

Improving classification based off-topic search detection via category relationships

Improving classification based off-topic search detection via category relationships

Region-based association measures for ranking mined news relations

News Relation Discovery Based on Association Rule Mining with Combining Factors

Contact Info

Product

Resources

About