AbstractIn courses in Logic at universities, some attention is always paid to translating natural language into a logical language, mostly proposition logic or predicate logic. To practice with this matter, students are presented several sentences which they have to translate. For many students, this is not at all easy and supporting feedback is needed badly.
When these exercises occur in a classroom context, it is thanks to the human flexibility of the teacher that every possible question can be answered. However, in an automated intelligent training environment it is difficult to provide feedback on any errors and on the questions of the students. This has two reasons: the complexity of natural language, and the process of translating itself. This is an activity that is difficult to dismember. Does it consist of a number of steps, and if so, what is the order in which they should be carried
Because of these reasons, intelligent learning environments have problems with providing feedback on translation exercises to students. Sometimes, it is rather limited, not much more than a judgement of right or wrong. Sometimes, feedback can also be excessively detailed. This is the case when the tutoring system assumes that translation consists of all kinds of intermediate steps, each of which has its own complexity and rules.
My goal is to design an intelligent learning environment for translation exercises of natural language in first-order predicates to specify logic in which translation does not have to be done in a complex series of steps and in which, in an automated way, errors made by the students as well as possible questions, can be addressed. To perform both tasks, error
detection and giving appropriate hints, this tutoring system is based on the theoretical concept
of “constraint”. This concept has successfully been used as a learning environment for
creating SQL queries, both to provide feedback and to judge the solutions of the students.
The scientific contribution of this thesis is to research whether the concept of “constraint”
can be usefully deployed in translation exercises.
To see whether the tutoring system developed did what it was intended to do, to be
an intelligent learning environment for translation exercises, the system has been made
available to students of the Open University. They worked with this constraint-based tutor
and their opinions as well as their interaction with this system, have been collected. The
tutor was appreciated well, although the feedback it gave sometimes missed the central
point. In those cases, the feedback of the system was difficult to understand. Given that
this is a novel approach, more research is needed on how it can be done present an extra
layer of feedback on translation exercises in situations where that looks necessary.
|Date of Award||9 Dec 2022|
|Supervisor||Josje Lodder (Examiner) & Bastiaan Heeren (Co-assessor)|
- Predicate logic
- Intelligent tutoring system
- Master Computer Science