Intelligent information triage

Macskassy, Sofus A.; Provost, Foster

doi:10.1145/383952.384015

Cited by 15 publications

(9 citation statements)

References 25 publications

(32 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Intuitively, this framework works well, and can be solved by combining some of the existing techniques [1,2,4,7,9,12,16,20,21,17,27,28,30,31,32,33,18,34,36,37,38]. Unfortunately, we claim that this framework is ineffective to solve our problem.…”

Section: Introductionmentioning

confidence: 92%

Time-dependent event hierarchy construction

Fung

Liu

et al. 2007

Proceedings of the 13th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

In this paper, an algorithm called Time Driven Documents-partition (TDD) is proposed to construct an event hierarchy in a text corpus based on a given query. Specifically, assume that a query contains only one feature -Election. Election is directly related to the events such as 2006 US Midterm Elections Campaign, 2004 US Presidential Election Campaign and 2004 Taiwan Presidential Election Campaign, where these events may further be divided into several smaller events (e.g. the 2006 US Midterm Elections Campaign can be broken down into events such as campaign for vote, election results and the resignation of Donald H. Rumsfeld). As such, an event hierarchy is resulted. Our proposed algorithm, TDD, tackles the problem by three major steps: (1) Identify the features that are related to the query according to both the timestamps and the contents of the documents. The features identified are regarded as bursty features; (2) Extract the documents that are highly related to the bursty features based on time; (3) Partition the extracted documents to form events and organize them in a hierarchical structure. To the best of our knowledge, there is little works targeting for constructing a feature-based event hierarchy for a text corpus. Practically, event hierarchies can assist us to efficiently locate our target information in a text corpus easily. Again, assume that Election is used for a query. Without an event hierarchy, it is very difficult to identify what are the major events related to it, when do these events happened, as well as the features and the news articles that are related to each of these events. We have archived two-year news articles to evaluate the feasibility of TDD. The encouraging results indicated that TDD is practically sound and highly effective.

show abstract

Section: Introductionmentioning

confidence: 92%

Time-dependent event hierarchy construction

Fung

Liu

et al. 2007

Proceedings of the 13th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

View full text Add to dashboard Cite

show abstract

“…Buchanan and Loizides () and Loizides and Buchanan () mention that (a) the scientific understanding of document triage is limited, (b) the research on triage is fragmented, and (c) triage has attracted little attention. Existing studies of triage include text classifiers for prioritization (Macskassy & Provost, ), display configurations (Bae et al, ), user activity logging (Badi et al, ; Bae et al, ), paper versus electronic documents (Buchanan & Loizides, ), enhancing section headings (Buchanan & Owen, ), visual search patterns (Loizides & Buchanan, ), and tag clouds (Maiya, Thompson, Loaiza‐Lemos, & Rolfe, ). Aside from the latter, studies use uniform document formats and small information spaces (e.g., 200 documents).…”

Section: Related Workmentioning

confidence: 99%

Collaborative exploratory search for information filtering and large‐scale information triage

Herceg

Allison

Belvin

et al. 2017

Asso for Info Science & Tech

View full text Add to dashboard Cite

Modern information seekers face dynamic streams of large-scale heterogeneous data that are both intimidating and overwhelming. They need a strategy to filter this barrage of massive data sets, and to find all of the information responding to their information needs, despite the pressures imposed by schedules and budgets. In this applied research, we present an exploratory search strategy that allows professional information seekers to efficiently and effectively triage all of the data. We demonstrate that exploratory search is particularly useful for information filtering and large-scale information triage, regardless of the language of the data, and regardless of the particular industry, whether finance, medical, business, government, information technology, news, or legal. Our strategy reduces a dauntingly large volume of information into a manageable, high-precision data set, suitable for focused reading. This strategy is interdisciplinary, integrating concepts from information filtering, information triage, and exploratory search. Key aspects include advanced search software, interdisciplinary paired search, asynchronous collaborative search, attention to linguistic phenomena, and aggregated search results in the form of a search matrix or search grid. We present the positive results of a taskoriented evaluation in a real-world setting, discuss these results from a qualitative perspective, and share future research areas.

show abstract

“…Receiver-Operator Characteristic (ROC) analysis is an evaluation technique used in signal detection theory, which in recent years has seen an increasing use for diagnostic, machine-learning, and information-retrieval systems (Swets, 1988;Provost & Fawcett, 1997;Ng & Kantor, 2000;Provost & Fawcett, 2001;Macskassy et al, 2001). ROC graphs plot false-positive (FP) rates on the x-axis and true-positive (TP) rates on the y-axis.…”

Section: Motivationmentioning

confidence: 99%

Confidence Bands for ROC Curves

Macskassy

Provost

Littman³

2003

Self Cite

View full text Add to dashboard Cite

In this paper we study techniques for generating and evaluating confidence bands on ROC curves. ROC curve evaluation is rapidly becoming a commonly used evaluation metric in machine learning, although evaluating ROC curves has thus far been limited to studying the area under the curve (AUC) or generation of one-dimensional confidence intervals by freezing one variablethe false-positive rate, or threshold on the classification scoring function. Researchers in the medical field have long been using ROC curves and have many well-studied methods for analyzing such curves, including generating confidence intervals as well as simultaneous confidence bands. In this paper we introduce these techniques to the machine learning community and show their empirical fitness on the Covertype data set-a standard machine learning benchmark from the UCI repository. We show how some of these methods work remarkably well, others are too loose, and that existing machine learning methods for generation of 1-dimensional confidence intervals do not translate well to generation of simultanous bands-their bands are too tight.

show abstract

Intelligent information triage

Cited by 15 publications

References 25 publications

Time-dependent event hierarchy construction

Time-dependent event hierarchy construction

Collaborative exploratory search for information filtering and large‐scale information triage

Confidence Bands for ROC Curves

Contact Info

Product

Resources

About