Research And Application Of Recommendation Algorithm In The Artwork Field

The traditional art search method is to use search engine similar to Baidu,Google and other information based on the user's input of the art keyword,artwork name,author and other information,thus useful artwork information.The amount of data is getting larger and larger,people want to find their own goals,and the information useful to them is closely related.,The artwork keywords entered by the user in the search engine may be inaccurate or too long;at the same time,because of the difference in the content of the artwork information,some artworks cannot be accurately searched by the search engine,resulting in the accumulation of artworks perceived by the user No.With the improvement of the intelligence level of modern information services,users cannot satisfy the traditional search mode of "people looking for art information",and even need a personalized push service for "art information looking for people".Here,a recommendation algorithm based on the combination of multi-modal features of artworks is designed to recommend similar artworks,and a complete similar artwork recommendation system is constructed.The specific work content is as follows:(1)Aiming at the actual use scenarios of people browsing artworks on the App,combining the three aspects of the artwork image features that are most interesting to people,the text features of the author and name combination,and the size features,the design is based on the artwork Item multi-mode Recommendation algorithm based on the combination of dynamic features.In each of the five art categories,a double grid search method is used to determine the best weight of the combination of artwork features.Using the artwork data in the actual project as a data set,the three characteristics of the artwork image feature,text feature,and size feature are tested in different combinations to calculate the accuracy of similar artwork recommendation under different combinations.,Recall rate and F(F-measure)value.Draw a chart according to the data results.From the result chart,it can be clearly seen that the method of using three mutually combined multimodal features of artwork image features,text features and size features to recommend artworks has a better recommendation effect.The feasibility and advantages of using multi-modal features to construct a recommendation system in the field of big data in artworks.(2)Based on the recommendation algorithm based on the combination of artwork multi-modal features,the Flask framework is used to build the server,and the i OS client App is created through the Xcode development tool to complete a similar artwork recommendation system based on the artwork Item multi-modal features.Finally,a detailed functional test was carried out on a similar art recommendation system based on the multimodal features of the art item.