Natural Language Rule Parsing for Schedule Automation


Creating work schedules for employees in an organization is a difficult problem since there are many different factors to consider in order for the schedule to be effective. Some examples of these factors are employee flexibility, work hour preferences, organization size and type, etc. Furthermore, available products may not be suitable for the specific needs of different companies due to various reasons. The partner organization of the proposed research develops web based software to facilitate users to create schedules. One of the objectives of the partner firm is to automate the scheduling process, which is currently done manually by the users of their systems. The proposed research would be to allow the users to input rules in natural language, which would then be translated into constraints to be applied on the scheduling algorithm. Thus, instead of creating schedules from scratch, the users should be able to automatically create a schedule by only inputting some rules, and then be able to adjust the schedule according to their needs once it is generated by the algorithm. Also, the proposed algorithm should be able to utilize knowledge from previously inputted rules and generated constraints to gradually improve the rule generation process. The algorithm should also take into account parameters such as the generated rules that were previously accepted or discarded by the users and also their frequencies. Different techniques such as classification of rules, using pre-specified rule templates, use of common-sense knowledge, etc. should be explored.


Fahim Hasan
Faculty Supervisor: 
Dr. Fred Popowich
British Columbia