Abstract:

   Relevance Feedback (RF) is crucial for building a user profile which is a fundamental element of different intelligent systems such as information retrieval, information filtering, and personalization. RF is affected by a number of contextual factors such as mood, stress level, and sentimental state of the user. Covid-19 pandemic imposed dramatic changes to the user environment as well as the search context. This paper investigates user’s search behaviour to identify the differences in the behavior between the contexts before and during the Covid-19 pandemic. This can be practically translated into identifying the differences in the relationship between the implicit feedback and the explicit relevance level between the two contexts. For this purpose, we conducted three user studies (i) Pre-COVID-19, (ii) Mid COVID-19 and (iii) after Covid-19. A user study was conducted on the same group of users on the three user studies. The Pre-COVID-19 user study took place before the pandemic started and the Mid-COVID-19 user study took place three months after the beginning of the pandemic. After Covid-19 stage took place after 18 months of the pandemic. A linear regression model was developed for each user study using IBM-SPSS. The analysis showed a significant variation in the user behavior between the two studies due to the COVID-19 context and its impact on user search behaviour. Also, two new RF parameters in Mid-COVID-19 were shown to have a significant relationship with the explicit user interest which were Mouse Clicks and Page/Down strikes. Furthermore, the comparison between the two models showed that the second regression model achieved a higher accuracy level that is attributed to the common behavioral change imposed by the pandemic.