QQ客服 官方微博 反馈留言

	OPINAX：一个有效的产品属性挖掘系统
	郝博一 ; 夏云庆 ; 郑方 ; HAO Bo-yi ; XIA Yun-qing ; ZHENG Fang
	2010-07-15 ; 2010-07-15
会议名称	第四届全国信息检索与内容安全学术会议 ; CNKI
关键词	意见挖掘 未登录词 依存分析 Opinion mining out-of-vocabulary word dependency parsing TP391.3
其他题名	OPINAX:An Effective Product Attribute Mining System
中文摘要	产品属性抽取是产品意见挖掘的重要任务之一,直接影响着产品意见挖掘的性能。本文提出了一种基于语言依存分析和语料库统计相结合的未登录(OOV)产品属性挖掘算法。该算法基于一个小规模基本产品属性集,从依存分析结果中提取与已有属性相关的统计特征,从而实现从生语科中抽取OOV产品属性,并对OOV产品属性进行排队,将可信度较高的OOV产品属性优先推荐。本文对抽取算法和排序方法的正确率进行了实验评测,并对比了不同统计特征的有效性。实验结果证明,在排队后的前200个产品属性中能取得87.5%的抽取正确率。; As a major task of the product opinion mining system,product attribute extraction influences performance of the system significantly.To find the out-of-the-vocabulary(OOV)product attributes,an effective attribute mining algorithm is proposed based on language dependency parsing and corpus statistical analysis.Based on a small set of standard product attributes,this algorithm applies the dependency parsing tool on review text to fund the potential OOV product attributes.Then statistical features extraeted from both the dependency parsing results and text content are used to filter out the invalid OOV product attributes and rank the attribute by measuring the confidence.Experiments are conducted to evaluate precision of OOV attribute extraction and effectiveness of the ranking method.Moreover, contribution of various features is also evaluated.Experiment results show that precision of OOV attribute extraction reaches 87.5% in the top 200 candidates.
语种	中文 ; 中文
内容类型	会议论文
源URL	[http://hdl.handle.net/123456789/69932]
专题	清华大学
推荐引用方式 GB/T 7714	郝博一,夏云庆,郑方,等. OPINAX：一个有效的产品属性挖掘系统[C]. 见:第四届全国信息检索与内容安全学术会议, CNKI.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。