| Mongolian history,as a cultural treasure passed down by the Chinese nation for thousands of years,is an important spiritual wealth of human civilization.As the main carrier for preserving and recording Mongolian history,Mongolian history books are of great significance to Mongolian history.In recent years,the development of recommendation technology has provided convenience for users to obtain personalized recommendation services.However,due to very little research on books that record ethnic culture and lacking of user reading records in the domain of Mongolian history books,it is impossible to directly use the single-domain recommendation algorithm to recommend books to users.In this thesis,the recommendation scenario that lacks user-related information in the domain of Mongolian history books is defined as the users ‘zero/low-resource’ recommendation scenario.In this recommendation scenario,this thesis introduces cross-domain recommendation technology and utilizes richer sentiment features.Then mining finegrained domain feature information to establish content associations between domains and building cross-domain recommendation algorithms.Finally the recommendation problem in the domain of Mongolian history books is solved.The main research contents are as follows:(1)With the background of movies and Mongolian history books,we use sentiment analysis and knowledge graph construction technology to extract aspectlevel sentiment features in movies and Mongolian history books and construct instance-level sentiment knowledge graph,which solves the problem of domain feature fusion in cross-domain recommendation and lays the foundation for subsequent research on cross-domain recommendation problems.(2)In this thesis,the sentiment knowledge graph is used as the content association between movies and Mongolian history books,and the Cross-domain Recommendation with Sentiment Knowledge Graph algorithm is constructed.Through three types of experiments and extended experimental results,it is proved that the algorithm achieved 83.61% recommendation accuracy in the domain of Mongolian history books,and achieved 73.6% cross-domain recommendation accuracy when extended to all categories of books,respectively achieving good recommendation results. |
