| Popis: |
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business Analytics With the prolific increase of structured and unstructured data, the need to have a tool that can provide the most relevant documents in an immense pool of data easily and efficiently has arisen. This issue can be found in the Scientific and Technical Information area, for example, in libraries with the steady intensification of the number of documents available. Although a Recommender System is a remarkable tool to help handle this information overload problem, it suffers gross inadequacy when low or no user information is available, thus termed the Cold Start Challenge. The current research aims to mitigate this predicament by using a Content-Base technique focused on text mining approaches, such as embedding models to exploit text-rich data. Experimental results show noteworthy adequacy in finding the most similar documents compared to a selected one in terms of syntactic and semantic meaning. These findings reveal that using embedding models is a promising approach to overcome the Cold Start Challenge of a Recommendation System when data is text rich. |
