Extracting Interests from Social Media to solve Cold Start in Personalized POI Recommendation Systems

Abstract

"The storage capacity of LBSN's have expanded mainly as result of an upsurge in visits and reviews. The prospective usefulness of such information in recommendation prompted the initiation of point of interest recommendation solutions development. The original solutions which were implemented began to incorporate a few primary important aspects, namely social, geological, content, and spatial, which have a bearing upon the suggestion of points of interest. Subsequently, these aspects integrated with contemporary recommendation methodologies such as collaborative filtering and content-based filtering. Nevertheless, current studies indicate that the utilization of media in social networking platforms has emerged as a promising avenue for capturing user engagement. Although the analysis of media constitutes a noteworthy effort in determining user preferences, its significance may be rendered futile in the absence of linguistic context. The study method has resulted in the identification of a gap. The method suggested in this document aims to bridge the existing gap by integrating linguistic context and image classification techniques to effectively deliver personalized suggestions. The suggested method aims to utilize social networking platforms as a foundation for identifying user preferences through the collection of likes on those platforms. The favorited content are divided into two components: media and text. The content will undergo text processing techniques such as sentiment detection, semantic evaluation and POS tagging to ascertain its intended meaning and the individual's subjective evaluation of it. The media will undergo classification in order to determine their respective representations. The findings are aggregated to enhance the precision and dependability of suggestions by offering a significant framework to the subject matter. The individual's preferences are subsequently inputted through a rating algorithm, which determines the most significant topics for the user. The algorithm will utilize these preferences to classify the locations it recommends to the individual."