Text Similarity Calculation Method Based on Optimized Cosine Distance
Jianqiang Zhang, Fangxu Wang, Futan Ma, Guoxing Song · 2022 International Conference on Electronics and Devices, Computational Science (ICEDCS) · 2022
In the text similarity calculation algorithm, the distance algorithm is widely used to measure the distance difference between two text vectors, and the cosine distance algorithm is fully verified. However, the measurement Angle of cosine distance focuses on the included Angle after text vectorization, that is, the direction of vector, without considering the speed of change of each vector. The improved algorithm proposed in this paper not only considers the direction dimension of the vector, but also studies the influence of the change of each dimension on the text similarity of cosine distance measure. Experiments show that compared with traditional cosine similarity, the improved algorithm in this paper has obvious improvement in accuracy and efficiency.