Disentangle Interactive Behaviour Representations According to Different Factors

Description: Zhang et al. [1] learnt mouse and keyboard behaviour representations. Such representations, in the embedding space, contains rich information on users, interactive tasks and user interfaces. Therefore, these representations are promising to be used to disentangle and understand how these different factors influence the behaviour.

Goal: Apply data analyses (statistics and/or deep learning approaches), visulaisations on the given interactive behaviour representations to understand if and how different factors influence interactive behaviour in the embedding space.

Supervisor: Guanhua Zhang

Distribution: 30% Literature, 60% Data analysis, 10% Visualisation

Requirements: Strong programming and mathematical skills.

Literature:

[1] Zhang et al. 2023. Exploring Natural Language Processing Methods for Interactive Behaviour Modelling. IFIP INTERACT'23.

[2] Han et al. 2020. Modelling user behavior dynamics with embeddings. Proceedings of the 29th ACM International Conference on Information & Knowledge Management.

[3] Ridgeway et al. 2018. Learning deep disentangled embeddings with the f-statistic loss. Advances in neural information processing systems .

[4] Miladinović et al. 2019. Disentangled state space representations. arXiv preprint arXiv:1906.03255.