Learning to Identify Educational Materials

International Conference RANLP 2009 - Borovets, Bulgaria, pages 123–127 Learning to Identify Educational Materials Samer Hassan and Rada Mihalcea University of North Texas samer@unt.edu, rada@cs.unt.edu Abstract In this paper, we explore the task of automati- cally identifying educational materials, by clas- sifying documents with respect to their educa- tive value. Through experiments carried out on a data set of manually annotated documents, we show that the generally accepted notion of a learning object’s “educativeness” is indeed a property that can be reliably assigned through automatic classification. Keywords learning objects, educational applications, text classification 1 Introduction With the rapid growth of the amount of information available online and elsewhere, it becomes increasingly difficult to identify documents that satisfy the user needs. Current search engines target broad coverage of information, at the cost of providing limited support for well defined verticals. In particular, an increasingly large number of users, consisting primarily of students, instructors and self- taught learners, are often seeking educational materi- als online, to use as standalone instructional materials or to supplement existing class resources. The typi- cal solution is to either refer to existing collections of learning materials, which often lack breadth of cov- erage, or to search the Web using one of the current search engines, which frequently lead to many irrele- vant results. For example, as shown later in Section 3, from the top 50 documents returned by a search per- formed on a major search engine1 for the query “tree data structure,” only four were found to be strongly educative, while as many as 29 documents were found to be non-educative. In this paper, we address the task of automatically identifying educational materials. We formulate the task as a text categorization problem, and try to au- tomatically classify the “educativeness” of a document (defined as a property that reflects the educative value of a document). Through annotation experiments car- ried out on a data set of materials from the domain of computer science, we show that the educativeness of a document is a property that can be reliably assigned by human judges. We also identify several features characteristic to educational resources, which can be used to identify the educativeness of a document. We 1 Throughout our experiments, we conduct our searches using the Google search engine. perform a number of classification experiments, and show that the document educativeness can be learned and automatically assigned. 2 Background A learning object is formally defined as “any entity, digital or non-digital, that may be used for learning, education or training” [2], or “any digital resource that can be reused to support learning” [12]. The idea that a document can have an educative property is widely accepted in the growing body of work dedicated to learning objects. Learning object repositories (e.g., [6, 8]) target improved access to learning materials through ”sharing and reuse,” by providing a common interface to entire collections of learning materials that can be shared among students and instructors and can be reused across courses and disciplines. …

Content truncated to 500 words