Research on the Identification of Scientific Research Project Misconduct Based on Deep Convolutional Neural Network

Science and technology are the primary productive forces, and scientific and technological innovation capability has become a decisive factor in the competition of comprehensive national strength. Scientific research projects are important carriers of scientific and technological innovation. With the implementation of the national innovation-driven development strategy, the number and funding of scientific research projects have increased significantly. The problem is that scientific research project misconduct (SRPM) emerges in endlessly. This paper studies the feature extraction method for misconduct of scientific research projects in large data environment, explores and proposes a sentence representation model and matching architecture based on deep convolution neural network (DCNN), and carries out effective training to learn the expression of sentence matching at different levels, so as to improve the expressive ability of the model, thereby improving the accuracy and accuracy of text similarity calculation.