An image-text fusion book recommendation method based on machine learning

Abstract

The invention relates to the technical field of library retrieval, in particular to an image-text fusion book recommendation method based on machine learning, which comprises the following steps of collecting and preprocessing the book related data, collecting the book image-text data from a network, and preprocessing the book image-text data; extracting the book picture features, training 1260000pictures in ImageNet 2012 by utilizing a DCNN and a VGG-16 deep convolutional neural network so as to obtain a relatively accurate training weight, extracting the picture features by utilizing the training weight, and performing dimensionality reduction on the picture features; extracting the book text features, using RNN and Word2Vec frameworks to convert a text into a vector consistent with animage vector in latitude, fusing the image features and the text features, and designing a linear integration method to fuse the image text vectors; realizing the recommendation, using a cosine similarity method for measuring the recommendation, calculating a classification threshold value, and recommending by combining a traditional item-based collaborative filtering recommendation method.

Description

technical field [0001] The invention relates to the technical field of library retrieval, in particular to a method for recommending books based on machine learning-based image-text fusion. Background technique [0002] Nowadays, the development of computer network technology is becoming more and more widespread. The traditional library model cannot meet the needs of the public, which makes the library innovate and reform all aspects of software and hardware under the new situation and develop rapidly. As an indispensable and important resource in people's life, information resources have shown unprecedented growth. The increase of service organizations has made the acquisition of information more extensive. People's needs for knowledge are more diverse and urgent. People's knowledge acquisition methods is always changing. As a result, the growth of information resources and the difficulty of information utilization are becoming more and more obvious. Therefore, libraries m...

AI Technical Summary

Problems solved by technology

[0004] Because the traditional university library uses the collection of documents to serve teachers, students and scientific research, it is time-consuming and labor-intensive to search and screen, and consumes a lot of manpower

Images