中文摘要：

关系抽取被广泛应用在问答系统、社会网络分析、本体构建等领域。将特征空间归纳为浅层分析和深层分析，进一步划分后者为句法结构和依存结构，总结了对每类特征空间的应用及扩展；分析了典型的计算相似度的核函数，归纳了对典型核函数的扩展方式；对已有相关研究的结果进行评价，分析了典型核的优缺点。最后基于上述优缺点分析，总结了现有研究的不足，指出了基于核的关系抽取的下一步研究方向。

英文摘要：

Relation extraction has been widely applied in many domains such as QA, social network, ontology construction. This paper categorized feature spaces into shallow parsing and deep parsing, and further divided deep parsing into the constitu- ent structure and dependency structure. It analyzed the application and extension of the above mentioned feature space classes subsequently. In terms of kernel function, this paper elaborated some representative kernel functions and summarized the ex- tended methods. Then, it concluded the performance of some kernels besides pros and cons regarding the representative ker- nels. Finally. this paper concluded the limitation of current researches and indicated some promising research directions.