October 05, 2018 – /,
Recommendation systems, in general, aim to reduce this burden of information overload by suggesting items of interest to a user. They imitate the human way of thinking by choosing as other like-minded people have chosen before when a decisive rst-hand knowledge is absent. Accordingly, they adapt the recommendation to the user's interest when additional meta-information is available.
An increasing number of companies and e-publishers are utilizing recommender systems in order to maximize the user's satisfaction and thus increase their return of investment (ROI). This is realized by leveraging the collected information about the user's interactions, preferences, and purchase history.
Traditional recommender systems, which aim to recommend documents, rely on statistical information only (collaborative filtering approaches) or on meta-information describing the user profile, the documents as well as their relations. They poorly consider the text semantics and mostly suffer from homophily by recommending documents which are similar to the documents already acquired by the user.
In this thessis, we want to design a query-based recommender system for the recommendation of textual documents. The system relies on Latent Semantic Model (CLSM) to implicitly measure the semantic similarity between the user’s interest and available documents. Taking into account the user’s previous activities, knowledge and current interests. The system represents the user’s history as a query of keywords. A graph-database will be used to prject the sematnic realions betweemn the documents data set.
In this work, we aim to develop/adapt machine learning methods (mainly Deep Learning) to improve the state-of-the-art in contnet-based recommender systems. This includes:
The written report must contain an introduction to the topic and provide an overview of related work. Furthermore, the designed and implemented methods must be described and discussed.
• Good programming skills in mind. A high level language mainly python
• Helpful: Previous experience in Natural Language Processing and Machine Learning
Beginning and duration
Immediately, duration 3-6 months (depending on the course)
Keywords: Recommeder System, Word Embeddings,Graph-database, Semantic-based Recommender, Ontology-based Recommender
Research Area(s): Knowledge & Educational Technologies