An Intelligent Travel Recommender System by Mining Behavioral Attributes From Online Travelogues in Malayalam – A Low Resourced Language

Abstract

Language technology involves various language processing tools and techniques which significantly contribute to Natural Language Processing (NLP). Among NLP, natural language text and speech processing are two emerging segments that require huge attention from research. Regional language processing with the advent of Artificial Intelligence brings umpteen opportunities, especially in the Indian context as many languages were spoken in different parts of the Country. A Recommender Model in the Malayalam language in Travel and tourism domain using unsupervised machine learning techniques is the intention behind this paper. Malayalam is a low-resource and highly inflected language that possesses a greater chance for ambiguity. Data sharing online platforms and social media are used as data collection sources, where the availability is still limited and challenging, which may cause scarcity of data. The works propose various methodologies to generate a custom-made scraping model from the social media written in the Malayalam Language and its preprocessing. A deep-level Travelogue Tagger has been specially constructed as part of the experiment. This paper proposes a recommender model based on traveler reviews using Collaborative filtering and Cosine similarity methods. The experiment succeeded with high precision.