Topic relevance systems in spoken language assessment

In recent years, an increased demand to learn English for business and education has led to growing interest in automatic spoken language assessment and teaching systems. With this shift to automated approaches it is important that systems reliably assess all aspects of a candidate's responses. The focus here is on one form of spoken language assessment: whether the response from the candidate is relevant to the prompt provided. This will be referred to as off-topic spoken response detection. The main techniques to investigate the development of topic relevance systems will be focused on state-of-the-art deep learning approaches, which aligns well with the goals of the EPSRC research area "Artificial intelligence technologies". Beyond binary classification of prompt-response pairs as on-topic or off-topic, the aim here is to offer a level of interpretability such that relevance models can offer some feedback to candidates by indicating the overlapping and unrelated topical areas of the prompts and responses. In order to further extend topic relevance systems, two-way human dialogue will be explored to determine the dynamic topical areas in any conversation; it would be further desirable to be able to generate a meaningful topical map summarisation for any exchange of spoken words.