Intelligent selection and retrieval of multiple time-oriented records

• Published in: Journal of intelligent information systems Vol. 35; no. 2; pp. 261 - 300
• Main Authors: Klimov, Denis, Shahar, Yuval, Taieb-Maimon, Meirav
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.10.2010; Springer Nature B.V
• Subjects: Accuracy; Analysis; Artificial Intelligence; C (programming language); Computer Science; Construction; Data Structures and Information Theory; Information retrieval; Information Storage and Retrieval; Intelligent systems; Intervals; IT in Business; Knowledge; Language; Lists; Medical; Modules; Natural Language Processing (NLP); Oncology; Ontology; Population; Records management; Studies; Temporal logic; Usability; Visualization; Intelligent user interface; Intelligent population-specification language; Temporal reasoning; Temporal abstraction; Multiple subject selection; Intelligent query language; Knowledge-based systems; Multiple subjects
• ISSN: 0925-9902; 1573-7675
• DOI: 10.1007/s10844-009-0100-0

Time-oriented domains with large volumes of time-stamped information, such as medicine, security information and finance, require useful, intuitive intelligent tools to process large amounts of time-oriented multiple-subject data from multiple sources. We designed and developed a new architecture, the VISualizatIon of Time-Oriented RecordS (VISITORS) system, which combines intelligent temporal analysis and information visualization techniques. The VISITORS system includes tools for intelligent selection, visualization, exploration, and analysis of raw time-oriented data and of derived (abstracted) concepts for multiple subject records. To derive meaningful interpretations from raw time-oriented data (known as temporal abstractions ), we use the knowledge-based temporal-abstraction method. A major task in the VISITORS system is the selection of the appropriate subset of the subject population on which to focus during the analysis. Underlying the VISITORS population-selection module is our ontology-based temporal-aggregation (OBTAIN) expression-specification language which we introduce in this study. The OBTAIN language was implemented by a graphical expression-specification module integrated within the VISITORS system. The module enables construction of three types of expressions supported by the language: Select Subjects, Select Time Intervals , and Get Subjects Data . These expressions retrieve a list of subjects, a list of relevant time intervals, and a list of time-oriented subjects’ data sets, respectively. In particular, the OBTAIN language enables population-specification, through the Select Subjects expression, by using an expressive set of time and value constraints. We describe the syntax and semantics of the OBTAIN language and of the expression-specification module. The OBTAIN expressions constructed by the expression-specification module, are computed by a temporal abstraction mediation framework that we have previously developed. To evaluate the expression-specification module, five clinicians and five medical informaticians defined ten expressions, using the expression-specification module, on a database of more than 1,000 oncology patients. After a brief training session, both user groups were able in a short time (mean = 3.3 ± 0.53 min) to construct ten complex expressions using the expression-specification module, with high accuracy (mean = 95.3 ± 4.5 on a predefined scale of 0 to 100). When grouped by time and value constraint subtypes, five groups of expressions emerged. Only one of the five groups (expressions using time-range constraints), led to a significantly lower accuracy of constructed expressions. The five groups of expressions could be clustered into four homogenous groups, ordered by increasing construction time of the expressions. A system usability scale questionnaire filled by the users demonstrated the expression-specification module to be usable (mean score for the overall group = 68), but the clinicians’ usability assessment (60.0) was significantly lower than that of the medical informaticians (76.1).