To assist Java programming educations, we have developed a Web-based Java Programming Learning Assistant System (JPLAS) that offers several types of programming exercises to cover different levels of students. Among them, the element fill-in-blank problem (EFP) was designed for novice students to study grammar and basic programming skills through filling in the blank elements in a Java source code. In general, an EFP instance can be generated by selecting an appropriate code in a textbook or a Web site, and applying the blank element selection algorithm to it. Since EFP is expected to involve the Java grammar, plenty of EFP instances have been prepared for JPLAS, where a student may select an EFP instance matching his/her level. In this paper, we propose a recommendation function of selecting a proper EFP instance to be solved next by considering the difficulty levels and grammar topics. We verify the effectiveness of our proposal through applications to 48 students in the Java programing class in Okayama University.