In order to solve the problem of sparse and missing example in the off-line handwritten signature authentication scenario, a handwritten signature authentication tree algorithm was proposed. The algorithm improves the perceptual hash algorithm based on the discrete Fourier transform, which is used to gather the subject information in the signature image, compress and generate the perceptual digest; then the cosine distance is used to calculate the distance between the perceptual digest vectors. According to the cosine distance matrix, an authentication tree composed by multiple nodes is constructed and the distance threshold is determined to authenticate the unknown data. The authentication result can be used for the self-renewal of the authentication tree. The experimental results show that the handwritten signature authentication tree algorithm has high accuracy and strong robustness. The false rejection rate and false acceptance rate on small datasets are significantly lower than the traditional machine learning algorithm, and its self-renewal mechanism can also cope with the style changes of signature handwriting very well.

In the Chinese medical question and answer task, question intention detection is a very important part. At present, the common intention detection methods mainly use the manually designed matching rules to find the problem features to detect the intention of the problem, but the use of a large amount of labor usually brings about problems such as high cost and poor versatility. A novel method of intention detection is proposed in this paper. First, the collected questions with different intention categories are used to construct intention feature words. Then, based on the BERT pre-training language model, a two-classification model of phrase similarity is constructed. By comparing the combination results of problem word segmentation and the similarity of intention feature words, the multi-classification problem of problem intention detection is transformed into a two-classification problem between multiple phrases. Then we can get the inclination of the question for each intention category, that is the intention category of the question. The experiment shows that the method based on the two-classification model of phrase similarity has better effect than the previous methods, and the F1 value in the test set reaches 90.1.

No abstract

No abstract

Given the increasing growth of the Web and consequently the growth of e-commerce, the application of recommendation systems becomes more and more extensive. A good recommendation algorithm can provide a better user experience. In the collaborative filtering algorithm recommendation system, many existing approaches to collaborative filtering can neither handle very large datasets nor easily deal with users who have very few ratings, this paper proposes an improved constrained Bayesian probability matrix factorization algorithm. The algorithm introduces a potential similarity constraint matrix for specific sparsely scored users to affect the user’s feature vector, and uses the Logistic function to express the nonlinear relationship of the potential factors, combined with the Markov chain Monte Carlo method for training. Finally, the data set is used for testing and comparative evaluation. This experiment prove that the algorithmic model can be efficiently trained using Markov chain Monte Carlo methods by applying them to the MovieLens and Netflix dataset. The experimental results show that the algorithm has better predictive performance and is suitable for solving the problem of sparse rating matrix of specific users.