Purpose - The purpose of this paper is to propose a novel citation recommendation system that inputs a text and recommends publications that should be cited by it. Its goal is to help researchers find the related works. Further, this paper seeks to explore the effect of using relational features, in addition to textual features, on the quality of recommended citations. Design/methodology/approach - In order to propose a novel citation recommendation system, first a new relational similarity measure is proposed for calculating the relatedness of two publications. Then, a recommendation algorithm is presented that uses both relational and textual features to compute the semantic distances of publications of a bibliographic dataset from the input text. Findings - The evaluation of the proposed system shows that combining relational features with textual features leads to better recommendations, in comparison with relying only on the textual features. It also demonstrates that citation context plays an important role among textual features. In addition, it is concluded that different relational features have different contributions in the proposed similarity measure. Originality/value - A new citation recommendation system is proposed which uses a novel semantic distance measure. This measure is based on textual similarities and a new relational similarity concept. The other contribution of this paper is that it sheds more light on the importance of citation context in citation recommendation, by providing more evidences through analysis of the results. In addition, a genetic algorithm is developed for assigning weights to the relational features in the similarity measure.

Keywords