In order to understand visitors’ behaviour onsite, we track them in the museum by logging implicit and explicit visitor preferences during their visits. However, the question is how to understand their choices and behaviour in order to use this for after visit recommendations (for both onsite and online explorations) and for extablishing a long term relationship between the museums and the vistors.

Explicit visitor preferences

Any preferences that are logged by visitors making explicit choices are explicit preferences. In fact, explicit preferences are choices that visitors made at the beginning of their visit like choosing the language of contents being shown in their visits, their preference in choosing a theme among the available themes, and their age range (e.g., child or adult).

Implicit visitor preferences 

Any preferences that are logged without making an explicit choice are considered as implicit preferences of visitors. Visitors’ dwell-time and their walk-through path are two examples of implicit preferences of visitors at museums.

Digital heritage object recommendation based on onsite preferences

We have extracted onsite preferences with the aim of doing digital collection or digital object recommendations for two different exhibitions at the Allard Pierson Museum and at the MUSEON. To make things clearer, we will focus on the recommendations based on the onsite logs in the Roman Gallery of the Allard Pierson Museum.

In order to understand visitors’ interests beyond their explicit preferences at the Check-In station, implicit preferences are extracted from the sensor logs that track visitors’ movements through the museum. The map below shows walk-though paths of visits at the Allard Pierson museum. The most important signal that is logged and interpreted as an estimation of a visitor’s interest in an object being shown at an exhibition is dwell-time of the visitor in front of the object. Dwell-time of visitors means how long they stayed in front of the point-of-interest (POI) and interact with it, which is proportional to the level of interest that they have in the object. Together with explicit preferences, this implicit signal is used in the object recommendation system that has been implemented in the meSch-software that is used for generating post-visit recommendations drawn from the vast Europeana collections portal.

Walk through behaviour: dominant transitions from check-in (C-in) to check-out (S). Numbers on the edges show count of users’ movements between point of interests.

Impact of connecting physical world to the digital world in cultural heritage visits

The main achievement of using onsite behavior of visitors to understand their interests is that visitors can get recommendations of some related objects, stories or even other museums to visit without making any effort to interact with the recommendation system themselves. So there is no need to fill in a profile, or give away your personal data or let someone know what you think your interests are.  They just do their normal visits in museums and get recommendations based on their behaviour there. Moreover, using mobile devices makes it very easy to expand this study and help museum visitors to plan for further exploartion of digital heritage online. Which, in fact, benefits not only the museum that integrated the meSch system, but all heritage institutions that contribute content to Euroepana!

We are certainly very excited about this development, but need to test and thoroughly integrate it a bit more before we can release it together with the meSch software platform.

References

[1] Seyyed Hadi Hashemi, Wim Hupperetz, Jaap Kamps, and Merel van der Vaart. Effects of Position and Time Bias on Understanding Onsite Users’ Behavior. To be published in CHIIR’16: International ACM SIGIR Conference on Human Information Interaction and Retrieval. 2016.

About the author

S. Hadi Hashemi is a Computer Science PhD student at the University of Amsterdam, and he is doing his PhD under the supervision of Dr. Jaap Kamps. He is working in the EU-meSch project since May 2014. His general research interests lie in Information Retrieval and Data Mining. Specifically, he is working on personalization, and contextual search and suggestion using search engine query logs and onsite human information interaction logs in the meSch project.