Expanding the Relevant Judgment List based on Classified Documents

An Information retrieval system is a software system that provides access to books, journals, and other documents. The IR system provides this access according to a specific user query to retrieve the intended book, document… etc. The better IR system is the system that produces better results in a standard testing medium that contains collections of documents, queries, and relevant judgment lists. Relevant judgment lists are used to match each document to its relevant query, and the result of the IR system is compared to this judgment list to evaluate its performance. Usually, judgment lists are built by specialists in the domains of the document subject. In this paper, we used a Machine learning algorithm to create a judgment list with the least human involvement, saving much time and effort. The algorithm that is used is K-nearest neighbor, and the method that was used is the inverse of the distance. The requirements for this method are a few human-matched documents forming a small relevant judgment list, which saves effort and time, as we said before. The experiments were done on a test collection: Medline, which was composed of 1033 documents with 30 queries. This experiment serves as a preliminary study conducted on a small test collection, providing a foundation for further refinement and validation on larger collections in future work.