Personalization in BERT with Adapter Modules and Topic Modelling

As a result of the widespread use of intelligent assistants, personalization in dialogue systems has become a hot topic in both research and industry. Typically, training such systems is computationally expensive, especially when using recent large language models. To address this challenge, we develop an approach to personalize dialogue systems using adapter layers and topic modelling. Our implementation enables the model to incorporate user-specific information, achieving promising results by training only a small fraction of parameters.