Searching for information on occupational accidents
Effective retrieval of the most relevant documents on the topic of interest from the Internet is difficult due to the large amount of information in all types of formats. Studies have been conducted on ways to improve information retrieval (IR). One approach to improve searches in large collections, such as the Web, is to take advantage of semantic representations in pre-existing relational databases that have been developed for explicit purposes besides supporting Internet searches in general. In an effort to enhance IR on the Internet, a prototype of a topic-oriented search agent, SAOA-1, was developed to use embedded semantics and domain-specific knowledge extracted from such a database. Activated when a set of retrieved keywords appears related to the topic of "occupational accidents", SAOA-1 constructs an alternative search query and pruned lists of suggested refinements by applying the search engine knowledge and the domain-specific knowledge and semantics extracted from a relational database. Information seekers could then use the alternative search query or refine it further with a modified search query developed by SAOA-1 based on its semantic representation of the topic of occupational accidents to complete context-sensitive pruning of the semantic neighborhood. An empirical study was conducted to evaluate the usefulness of SAOA-1 in assisting information seekers to retrieve relevant documents. Sixty participants were randomly assigned to one of two treatments: with or without the assistance of SAOA-1, to perform Internet searches. Prior to performing searches, each participant had to decide upon a topic based on two given articles addressing hand injuries in the workplace. The participants then performed searches and, when satisfied with the results, evaluated the relevance of the first forty documents of their final search to their research topics on a 1-to-5 scale Likert scale. It was hypothesized that the treatment type could have an overall effect on expected rating. Based on the data collected, the average expecting rating was statistically significant (p < 0.001) with a slight improvement of 10% due to the treatment. From a practical perspective, however, the size of the effect was modest. The findings suggest that a topic-oriented search agent might be useful in assisting information seekers to retrieve more relevant documents but suggest important directions for further evaluating methods and settings for taking advantage of such semantically-based search techniques.
School:The Ohio State University
School Location:USA - Ohio
Source Type:Master's Thesis
Keywords:information retrieval semantic representations
Date of Publication:01/01/2008