Entity Matching Analysis using SIF, RNN, Attention and Hybrid Methods for Research Article Similarity

When searching for scientific research articles, there is often an abundance of search results from various sources, which makes it difficult to determine the relevance between topics. This encourages the use of entity matching techniques and three deep learning methods: RNN, Attention, and Hybrid methods. We utilize a deep learning approach that combines RNN, Attention, and Hybrid models to improve entity matching capabilities. With this approach, a dataset of scientific articles based on the title, author, and keywords can be processed, and the RNN model can be used to learn the representation of those features separately. Furthermore, the Attention model is used to derive the relevant weights between pairs of features, while the hybrid model combines the results from the RNN and Attention models. In this context, the combination of RNN, Attention, and Hybrid methods can be a solution in determining relevance between scientific articles, improving efficiency, and effectiveness in research literature search. Based on the results of the analysis, the highest results were obtained in the hybrid model, with an F1 score accuracy of 51.18%.