SFP-Rank: significant frequent pattern analysis for effective ranking

Song, Yuanfeng; Ng, Wilfred; Leung, Kenneth Wai-Ting; Fang, Qiong

doi:10.1007/s10115-014-0738-y

Cited by 6 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The study reported evaluation results in a different metric than the metrics chosen for this meta-analysis (Yu and Joachims, 2009;Thuy et al, 2009;Pahikkala et al, 2009;Kersting and Xu, 2009;Mohan et al, 2011) 6. The study reported a higher performance on baseline methods than official benchmark runs (Dubey et al, 2009;Banerjee et al, 2009;Peng et al, 2010a;Song et al, 2014;Bian et al, 2010;Bian, 2010;Carvalho et al, 2008;Acharyya et al, 2012;Peng et al, 2010a;Tran and Pham, 2012;Asadi, 2013) 7. The study did not report any baseline performance that allowed us to check validity of the results (Chakrabarti et al, 2008;Wang et al, 2012b;Buffoni et al, 2011).…”

Section: Literature Selectionmentioning

confidence: 99%

A cross-benchmark comparison of 87 learning to rank methods

Tax

Bockting²,

Hiemstra

2015

Information Processing & Management

View full text Add to dashboard Cite

Learning to rank is an increasingly important scientific field that comprises the use of machine learning for the ranking task. New learning to rank methods are generally evaluated on benchmark test collections. However, comparison of learning to rank methods based on evaluation results is hindered by nonexistence of a standard set of evaluation benchmark collections. In this paper we propose a way to compare learning to rank methods based on a sparse set of evaluation results on a set of benchmark datasets. Our comparison methodology consists of two components: 1) Normalized Winning Number, which gives insight in the ranking accuracy of the learning to rank method, and 2) Ideal Winning Number, which gives insight in the degree of certainty concerning its ranking accuracy. Evaluation results of 87 learning to rank methods on 20 wellknown benchmark datasets are collected through a structured literature search. ListNet, SmoothRank, FenchelRank, FSMRank, LRUF and LARF are Pareto optimal learning to rank methods in the Normalized Winning Number and Ideal Winning Number dimensions, listed in increasing order of Normalized Winning Number and decreasing order of Ideal Winning Number.

show abstract

Section: Literature Selectionmentioning

confidence: 99%

A cross-benchmark comparison of 87 learning to rank methods

Tax

Bockting²,

Hiemstra

2015

Information Processing & Management

View full text Add to dashboard Cite

show abstract

“…The greater the resemblance to the reference summary, the great results will be. The following graph shows the summary vise comparison of all the summaries generated for 48 folders for both Topic Signatures and Density Peak clustering algorithm [6]. An experimental result on the dataset shows that different optimizing techniques can be used to achieve better results for summarization tasks.…”

Section: Discussion and Outcomesmentioning

confidence: 99%

“…Several models are proposed to rate the most appropriate and significant sentence to generate the best summaries. Most previous methods have used clustering-based approaches where the researchers take clusters and rank them in two independent steps mentioned by [5] and Song et al [6]. The rest of the researchers manage the process of clustering in a sharing manner, assuming that clustering mends the sentence ranking properly.…”

Section: Related Workmentioning

confidence: 99%

What have you read? based Multi-Document Summarization

Irum

Nasir

Jalil

2022

IJIST

View full text Add to dashboard Cite

Due to the tremendous amount of data available today, extracting essential information from such a large volume of data is quite tough. Particularly in the case of text documents, which need a significant amount of time from the user to read the material and extract useful information. The major problem is identifying the user's relevant documents, removing the most significant pieces of information, determining document relevancy, excluding extraneous information, reducing details, and generating a compact, consistent report. For all these issues, we proposed a novel technique that solves the problem of extracting important information from a huge amount of text data and using previously read documents to generate summaries of new documents. Our technique is more focused on extracting topics (also known as topic signatures) from the previously read documents and then selecting the sentences that are more relevant to these topics based on update summary generation. Besides this, the concept of overlapping value is used that digs out the meaningful words and word similarities. Another thing that makes our work better is the Dice Coefficient which measures the intersection of words between document sets and helps to eliminate redundancy. The summary generated is based on more diverse and highly representative sentences with an average length. Empirically, we have observed that our proposed novel technique performed better with baseline competitors on the real-world TAC2008 dataset.

show abstract