论文部分内容阅读
天网知名度系统是根据用户预定信息提供个性化检索的信息服务系统。该文提出了一种基于概率模型的名人网页相关度评价模型。改进O kap i BM 25公式,引入HTM L标记权重系数针对不同领域名人特点引入名人属性权重系数。分别采用伪反馈和用户反馈两种方法进行相关反馈,实现对评价模型中权重参数的自动优化。实验表明,该模型有效地提高了系统相关度评价质量,并且发现用户反馈的效果受实体属性信息词数影响,属性信息越丰富反馈后性能提高的概率越大。
Skynet visibility system is based on user booking information to provide personalized search information service system. This paper presents a model of celebrity webpage relevance evaluation based on probability model. The O kap i BM 25 formula is improved and the HTM L Marker Weight Coefficient is introduced to introduce the celebrity attribute weight coefficient according to the celebrity characteristics in different fields. Relevant feedbacks are respectively made by using two methods: pseudo-feedback and user feedback, so as to automatically optimize the weight parameters in the evaluation model. Experiments show that this model effectively improves the quality of system relevance evaluation and finds that the effect of user feedback is affected by the number of entity attribute information words. The more the attribute information is, the higher the probability of performance feedback is.