This project will create intelligent tutoring tools that generate mathematical word problems and worked solutions using Natural Language Generation technology. Problems and solutions will be delivered through interactive multimodal interfaces that will be evaluated with school children and maths tutors.
Mathematical word problems are problems wrapped up in narratives, e.g., "Debbie deals 46 playing cards to Thomas and herself. Now, Debbie has 14 more than Thomas. How many do they each have?"
Solutions will be generated as hints, starting with the most general: "Give Debby her 14 extra playing cards and then share the rest equally between them." and gradually getting more and more specific: "How many are left when we give Debby her 14?" and "46 - 14 = ?" then "32 playing cards are left." and "Share them equally." and so on.
Almost half of the adults in Europe have poor numeracy. They have problems with simple sums in everyday situations such as calculating change in a shop. This makes it hard for them to get jobs and have a good quality of life. New techniques that we develop may help improve numeracy, thus contributing to a better society and economy.
The project will address the challenges of how to vary reading difficulty and complexity of encoding mathematics problems in language, especially how to create coherent stories, how to plan different problem solving strategies for each type of question, and how to present them in a structured manner.
This research will help to open the way for a new breed of sophisticated tutoring applications that exploit language technology for (1) posing questions and (2) providing hints that to help students work out and learn solving strategies.
Go to Maths Stories publications.
Go to Maths Stories Demo Page