| 研究実績の概要 |
As there is no clear definition about a user's "information need" in searching online teaching materials, our first job is to formalize the concept of what 'search' process means in an "Online Teaching Materials Information Retrieval System" or OTMIRS in short. So far we have figured out how users are going to perform the search process, and what the search results should be. On this principle, we have analyzed what and how teaching materials are to be organized and indexed in the system.
Basically, much like in other IR systems, users initially provide a few relevant keywords about the material one is looking for. Systems, however, would have no idea whether one's particular "information need" is a full classroom lecture or just a particular section. However, although a user may not have a definite "information need" in mind at the start of the query, one's knowledge may evolve over the time as one keeps interacting with the system.
For our OTMIRS system, we have figured out how to fragment different media of online teaching materials. For example, lecture slides can be fragmented into individual ones. Each single slide may be further fragmented into title, body with different levels of details, figures, tables etc. Similarly, a reference book may be fragmented into chapters, sections, subsections, paragraphs and sentences etc. An audio/video material may be fragmented into several shorter units automatically by using the inclusive information in the media such as ‘pause/silence’ or analyzing particular terms such as ‘next, by the way’ etc.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
3: やや遅れている
理由
As online teaching materials come in different digital format and media, our first job is to find a way of defining relationships among these media. Our idea at the start of the project was to develop a data model based on graph theory in order to represent semantic relationships among various sections of the several different materials. However, a graph model may quickly become cumbersome as soon as the data size becomes larger, which may search process inefficient. As an alternate approach, we are developing an idea to represent several sections of materials as different objects in a multi-dimensional space, where each object is placed in such a way that it represents its semantics as well as contextual importance.
|
| 今後の研究の推進方策 |
Based on our new approach to index online teaching materials, we are going to extend the idea of specifying users' search process in multiple dynamics. In particular, when a user provides query keywords, the system initially attempts to capture what the user's information needs are. Further, based on the initial results provided by the system, we expect users' knowledge to evolve so that they make further modifications in the querying process. Although these ideas already exist in some of the IR systems, we are going to develop a formal model that is particularly suitable for an Online Teaching Materials Information System.
|
