Recommendation of tourism resources supported by crowdsourcing

Resumo

In the last decades, travelling has changed dramatically due to the evolution and popularisation of ICT as well as mobile devices, namely, smartphones. Concretely, ICT have revolutionised the tourism industry as well as the tourist behaviour by allowing permanent access to Web-based platforms holding large amounts of crowdsourced information. Crowdsourcing has become an essential source of information for tourists and the tourism industry. Increasingly, tourists search on and contribute to tourism crowdsourcing platforms. Every day, large volumes of tourism-related knowledge accumulates as tourists leave their digital footprints in the form of searches, posts, shares, reviews or ratings in these platforms. This crowdsourced information classifies prior tourist experiences and influences the planning and decision making behaviour of future tourists. Although tourism crowdsourced information influences decision making, typically, a tourist cannot monitor or control his/her own crowdsourced footprint to enhance his/her options, due to the complexity of the diverse platforms and resources. The problem of selecting tourism information, from large and heterogeneous datasets, based on the tourist profile is complex and requires specific tools. Using recommendation systems it is possible to suggest tourism resources according to the tourist digital footprint, i.e., the tourist profile, using artificial intelligence methodologies to mine the crowdsourced information. In this PhD dissertation, we focus on the problem of the personalisation of tourism recommendations based on crowdsourced information. In this context, we have designed multiple recommendation approaches to suggest tourism resources, supporting the travel cycle. Concretely, our contributions address: (i) the impact of ICT in the tourist experience; (ii) the profiling of tourists and resources based on crowdsourced ratings, reviews and views; and (iii) the personalised recommendation of tourism resources, using off-line and on-line content-based and collaborative algorithms as well as post-filters.