There is an ever-increasing need to apply metadata to legacy electronic training material as well as to content currently under development. Metadata, or very simply data about data, provide an underlying description of training material. Metadata describe attributes of learning objects including, but certainly not limited to, the content itself, when it was created, who created it, and its intended purpose. This information can allow developers to search and find previously developed content in order to achieve a financial efficiency through updating or reusing existing content. Further, as the future vision of Navy training matures, metadata can help ensure that sailors receive the right training at the right time based on knowledge of an individual sailor's needs and applicable training material.
Metadata are comprised of both objective and subjective data elements. Objective elements are those that are relatively straightforward to identify. They include data such as the developer, the training title, or the revision number of the content. Subjective elements - arguably the more valuable data - more thoroughly describe the training content. However, they are subject to individual interpretation and present a potential time consuming and expensive component to generating metadata. It is very appealing, therefore, to apply automation to the process of generating metadata. Technologies are available to assist in this process. Most notably, the application of a machine learning technology, Latent Semantic Analysis (LSA), can assist in the very arduous task of identifying subjective metadata tags.
This paper will describe the use of LSA in automating the metadata tagging process. Further, results of a research effort examining the use of LSA for metadata tagging will be presented. The results of this study indicate that the most efficient and effective process of tagging electronic training content may be to allocate that function between both the human and the computer.