User Representations for Language Learning

In our increasingly interconnected global society, learning foreign languages is an important factor for social mobility, and
teaching and learning platforms for foreign language learners act as a great educational equaliser. These platforms are
becoming increasingly sophisticated and make use of automation to give tasks to the learner in ascending order of
difficulty for efficient learning. In order to be successful at this, these platforms must have a reliable representation of the
learner's abilities. In this research, we will explore creating such representations. There are a number of complexities
involved in capturing language ability. One is that unlike the STEM domain where skills can be quantised and placed in a
hierarchy, linguistic ability is more complex. Skills may overlap, the hierarchy is not clear and many skills are continuous
in nature. Another problem is that it is difficult to capture what people know, as they only what they exhibit that they
know. We will explore these problems in my research, motivated by psycholinguistics understanding of human language
acquisition. A further problem lies in the gap between language comprehension and language production. This project will explore
modelling writing proficiency and reading proficiency separately, calculating a mapping between the two. Such
investigations have the potential to advance our understanding of language acquisition and could have direct
applications for modern language learning platforms.