elective courses is often a challenge for students due to complex considerations. They may face some problems such as being unable to obtain information or consult with programme advisors and peers. Therefore, we developed an educational chatbot for helping students to solve the problems on choosing elective courses by providing information, students' opinions and recommendations at the same time and any time of the day. The aims and objectives have been satisfied. All functions of EASElective are completed. After evaluating our solution, the findings of the system performance and user perception evaluation shows that the performance and functionality are positive.

EASElective has potential for course advising based on the result of evaluation. The majority of test subjects thought that our solution is more preferable and convenient than the existing solutions. According to the result of accuracy test, it was also found that nearly 90% of the information provided by the system is correct. Also, based on the result of user perception evaluation, the testing subjects thought that the system is easy to use, the dialogue is clear to understand and the response is accurate after experiencing our chatbot. It was also found that the average score of final year students is higher than non-final year students, which means our chatbot can help to solve the problems of making decisions on choosing elective courses. As they have already understood the rules of course selection and the characteristics of the courses that they are learning, this result reflects our chatbot can provide accurate and reliable information and recommendations.

Some limitations of our solution has been figured out. The major function of our system is to provide comprehensive information but it is difficult to collect useful and accurate students opinions to enrich our database. Although there are pre-defined routines to ask for users' opinions when the user has completed a course, we cannot ensure the user would answer these questions. On the other hand, the system cannot collect useful information for further use if some users respond meaningless opinions.

For future works, a more effective way should be investigated for students to obtain official information through the chatbot interface, due to the findings show that students found the course website and booklet more preferred to use. Also, more routines should be designed to handle more conversation situations, which can answer more questions. Second, more functions should be developed for analyzing possible selections and providing recommendations. The recommendation should be provided based on the students' preferences. The chatbot should ask more about their preferences and constraints in the conversation. If the recommendation is based on the students' preferences, users can obtain a more comprehensive advising.