An N-gram-based Information Retrieval Approach for Surveys on Scientific Articles

Humans are constantly searching for knowledge. This quest for knowledge has pushed back the boundaries of science. As a result, new scientific contributions are published daily in a variety of fields. However, it is not easy for a novice researcher to visualize all existing scientific contributions to a specific research problem in a short period of time. This study proposes an approach for extracting useful information from the metadata of scientific documents. Then, the design of an intelligent search system exploits the metadata contained in scholarly documents to provide an overview of scientific contributions to a research problem. The proposed model uses a new similarity measure based on the extraction of n-grams from the metadata of scientific articles. The model offers each user the possibility of visualizing the results of scientific contributions proposed by researchers in the form of a graph. Experiments carried out on a dataset of 126k data show that the model we propose achieves an overall precision of 0.89, a recall of 0.84 and an F1-score of 0.86. This shows that the model can refine the search to provide scientific contributions that have a direct correlation with a user’s need.