Dissertation Information for Ruth K. Maloney MurrayNAME:
DEGREE:
DISCIPLINE:
SCHOOL: ADVISORS: COMMITTEE MEMBERS: MPACT Status: Incomplete - Inspected Title: AN ANALYSIS OF GROUP PROFILING FOR SDI SYSTEMS IN INDUSTRIAL COMMUNITIES Abstract: The objectives of this research were to compare the effectiveness of: (1) three methods of determining membership in a group for SDI profiling purposes; (2) title searching with titles-and-abstract –text searching. Group profiling.—those methods of group profiling under study were grouping by organizational structure, by project orientation, and by self-expressed interest. Individual profiles were constructed for 104 chemists from three industrial research laboratories and matched against current issues of machine-readable Basic Journal Abstracts (BJA); individual outputs were distributed to each user. Users within each company then were assigned to groups (a group, by definition, comprising two or more persons) for profiling purposes. Twenty-two organizational and 27 project groups, involving 100 and 79 users respectively, were created from information supplied by the participating companies. Twenty-one interest groups, involving 53 users, were formed by examining the terms listed in the individual profiles to determine commonality of interest. A profile for each group was constructed by first merging all terms in the individual profiles, then eliminating those terms requested by only one user within the group. Eliminating single-frequency terms resulted in complete elimination of nine group profiles. The outputs retrieved by the group profiles were compared with individual outputs of group members to determine the following measure: the number of citations retrieved by both the individual and group profile (Equality), citations retrieved by the individual profile but not by the group profile (Loss), and citations retrieved by the group profile by not by the individual profile (Noise). These three measures were combined by a formula derived for the purpose of determining a performance value for each group profile for each individual. Provision for a weighing factor was included in the formula so that one could vary the relative importance that may be attached to noise when compared with loss as evaluative measures of profile performance. The Exact Binomial Test for significance was applied to thirty-six different comparisons of the performance values of the three methods of grouping to determine if one was significantly better statistically than another. Title versus title-and-abstract-text searching.—the BJA program searches both a citation and an abstract for each unit, but lists only citations at the output stage. Each individual user’s profile terms were matched against the citations in his individual outputs to determine the number of citations that would have been retrieved if titles only had been searched. The relative effectiveness of title searching as compared with title-and-abstract-text searching for the total group of users was computed as a ratio of title hits to total hits. Conclusions.—(1) Grouping by organizational structure was the least effective method, based on the tests for statistical significance. (2) No statistically significant difference could be established between grouping by project assignment and grouping by self-expressed interest within individual companies; when users of all companies were combined, results of two of six comparisons indicated grouping by project was more effective than grouping by interest. (3) The percentage of retrieval of citations if only titles are search when compared within searching of titles and abstract text can be expected to fall between 18.2% and 35.8% with a 95% confidence level; the mean of this population was 27.0%. (4) One can expect a higher level in the percentage of retrieval by titles only if only a single term is required for a match than in searches requiring more than one term to be present to constitute a match (39% compared with 16.8% in this study). |
MPACT Scores for Ruth K. Maloney MurrayA = 0 Advisors and Advisees Graphgenerating graph, please reload |
Students under Ruth K. Maloney Murray
ADVISEES:
- None
COMMITTEESHIPS:
- None