摘要
目前的上下文处理方法对语义技术支持不足,导致语义搜索在精准度和智能化上存在局限。提出一种基于BDI(BeliefDesire-Intention)的语义搜索方法,对动态上下文进行语义描述,利用BDI机制优化搜索结果。该方法将用户查询、动态上下文与BDI框架的Belief融合,通过产品特征语义、业务语义推理上下文特征和搜索目标特征作为Desire,通过基于Min Hash的语义相似度计算进行候选Intention排序。结合项目需求,在JADEX开源BDI Agent环境下,开发了基于上述语义搜索的旅游产品搜索系统。该系统通过ID3决策树提取用户当前行程的上下文特征实现了语义搜索功能,实验表明该方法搜索准确率相比仅使用语义相似度搜索方法明显提高,能够满足旅游产品搜索的要求。
Current context processing method does not support semantic technology very well, this leads to the limits of semantic search inboth accuracy and intelligence. In this paper, we propose a BDI-based semantic search approach, it makes semantic description on dynamiccontext and uses BDI mechanism to optimise search results. The approach integrates user, s query and dynamic contexts with Belief in BDIframework, reasons the context features through products feature semantics and business semantics, and searches objects features as agent’ sDesire, finally it calculates the candidate Intention order through MinHash-based semantic similarity. In combination with the requirement ofproject, we develop the search system of tourism products in environment of open source BDI Agent of JADEX engine, the system is based onthe above mentioned semantic search. Through ID3 decision tree the system extracts context features of current itinerary of users to implementthe semantic search function. Experiment shows that the search accuracy of the approach improved significantly relative to the search methodsonly using semantic similarity, it can satisfy the demand of tourism products search.
作者
陈昊
李银胜
沈剑平
Chen Hao;Li Yinsheng;Shen Jianping(School of Software, Fudan University, Shanghai 201203 , China;National Engineering Lab of Ecommerce Technologies, Shanghai 201203 , China)
出处
《计算机应用与软件》
CSCD
2016年第8期30-35,共6页
Computer Applications and Software
基金
国家科技支撑计划项目(2012BAH18F06)
关键词
BDI
语义搜索
上下文
Belief-desire-intention (BDI)
Semantic search
Context
