Web mining to generate multimodal learning materials for children with special needs

Developing a Web mining scheme that takes plain text (for example, a paragraph of a story) as an input to generate multimodal learning materials from the Internet sources could be of great help to both teachers and children with special needs. This paper proposes a solution which at first extracts the keywords from the plain text provided using word co-occurrences method. The co-occurrence distribution between frequent terms and other terms in a sentence are used to define the relative importance of a term. Thus extracted keywords are used to search for the multimodal elements (texts, images and video clips) and other similar stories through search engine APIs (Application Programme Interface). After the search operation, URL (Uniform Resource Locator) filtering, domain filtering and Web content analysis based filtering methods are used to remove unwanted materials. The proposed scheme has been implemented, tested and verified by children through ethnographic method to demonstrate the merits and capabilities in generating multimodal learning. Results show that the proposed scheme offers on average of about 80% accuracy for learning material generation.