Personality Based Recommendation Systems

The TWIN Personality Based Recommender System

With the appearance of the more intuitive tools of content manipulation and administration, the model of the person's interaction with the Web has changed. The user and her interests and needs have become the main starting point of all actions, held by online intelligent applications appearing elsewhere. They exploit the «wisdom of the crowds» to get the broader view over the particular area of knowledge and give an appropriate advice to the end user.

In the real world the person is surrounded by other people almost all the time so there is always a potential possibility to obtain help. And when going online the process is imitated by special types of intelligent Web services known as recommender systems. Such systems automatically provide expert opinions for people who tend to rely on them when choosing from the large number of alternatives.

One of the domains in which the necessity of making a good choice is very important is travelling. People are faced with a high degree of uncertainty when choosing a place (hotel, restaurant) they have never been to, consequently they must rely on other travellers' reviews which sites such as TripAdvisor ( provide. However the volume of user-generated content (i.e. reviews) grows in size very fast which prevents users from considering of all the available variants and becomes a separate problem of appropriate knowledge extraction that can be solved by means of the recommender system.

In our research we address the problem of estimating the similarity between people. We make an assumption that it can be established by analyzing the context of the words people use. Accordingly, the occurrence of the particular words in the particular text reflects the personality of the author. This suggestion leads to the possibility of the text-based detection of a circle of «twin-minded» individuals whose choices (e.g. hotels reviewed in TripAdvisor) could be quite similar and thus could be recommended to each other.

Currently we are developing the «Tell me What I Need» (TWIN) Personality-based Recommender System. We apply the system in the travelling domain, to suggest hotels from the TripAdvisor site by filtering out reviews produced by people with like-minded views. We aim to identify key personality characteristics of users (here we consider the Big Five personality model) based on linguistic cues collected from the user-generated texts in order to create personality-based user profiles. This approach provides recommendations that rely on the factors independent in many ways from the user's preexisting attitudes in the hotels market and also avoids the subjective step of specifying explicit preferences.

You can learn more about the TWIN system here.

  • Synergy Centre at ITT Dublin
  • QQI Brand Logo
  • Member of the Technological University for Dublin Alliance
©2019 Institute of Technology Tallaght • Built by Digital Crew