Scholarlink

There is a wealth of digital information available for consumption today, including e-books, movies, videos, and articles. It is overwhelming to go through everything available and choose what to watch next. Therefore, digital media providers aim to take advantage of this misunderstanding and address it in order to boost user engagement, which will eventually result in increased revenues. In order to combat this information overload, content producers frequently use recommendation algorithms. The goal of online movie recommendation systems is to address the movie information boom and provide viewers with personalized recommendations. In the past, collaborative filtering—which makes use of user involvement with the media—or content-based filtering—which makes use of the available metadata for the films—have been the two main methods used by movie recommendation systems. Because they can combine different recommendation models and tackle the problems of data sparsity and could start to increase recommendation performance, knowledge graphs have recently shown to be extremely useful for recommender systems. However, the information about the user's properties is taken into account less than the information about the properties of the products in the current knowledge graph recommendation algorithms, which results in certain restrictions in the recommendation results. The focus of this research is on creating a synthetic method for movie recommendations. Another result of technological development is a hybrid method that combines the two systems. Our strategy, on the other hand, focuses only on content-based recommendations and improves them with a ranking algorithm based on measures for content similarity. To overcome the aforementioned difficulties, the suggested approach incorporates various information to learn the users' and objects' vector representations for top-N recommendations. We take the information about films from the Linked Open Data and use the knowledge representation learning approach to embed it into a single vector space together with information from real-world datasets used by recommender systems. A preliminary list of recommendations is created using additional calculations on these vector representations.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Context Aware Comprehensive Movie Recommendation System Using Semantic Word Embedding Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • K. Likitha,
  • N. Ram Harini,
  • K. Archita,
  • K. Ramkumar

摘要

There is a wealth of digital information available for consumption today, including e-books, movies, videos, and articles. It is overwhelming to go through everything available and choose what to watch next. Therefore, digital media providers aim to take advantage of this misunderstanding and address it in order to boost user engagement, which will eventually result in increased revenues. In order to combat this information overload, content producers frequently use recommendation algorithms. The goal of online movie recommendation systems is to address the movie information boom and provide viewers with personalized recommendations. In the past, collaborative filtering—which makes use of user involvement with the media—or content-based filtering—which makes use of the available metadata for the films—have been the two main methods used by movie recommendation systems. Because they can combine different recommendation models and tackle the problems of data sparsity and could start to increase recommendation performance, knowledge graphs have recently shown to be extremely useful for recommender systems. However, the information about the user's properties is taken into account less than the information about the properties of the products in the current knowledge graph recommendation algorithms, which results in certain restrictions in the recommendation results. The focus of this research is on creating a synthetic method for movie recommendations. Another result of technological development is a hybrid method that combines the two systems. Our strategy, on the other hand, focuses only on content-based recommendations and improves them with a ranking algorithm based on measures for content similarity. To overcome the aforementioned difficulties, the suggested approach incorporates various information to learn the users' and objects' vector representations for top-N recommendations. We take the information about films from the Linked Open Data and use the knowledge representation learning approach to embed it into a single vector space together with information from real-world datasets used by recommender systems. A preliminary list of recommendations is created using additional calculations on these vector representations.