Abstract: Consumption intention recognition in microblog is a new research topic in social media. However, existing work mainly focuses on identifying the commercial intention in microblog, literatures engaged in the recognition of consumption target are still sparse. Meanwhile, recognizing consumption target is the key to accurate commercial recommendation. Since the contents in microblog are usually short and informally using of terms, the results of dictionary matching and dependency analysis based methods are not satisfied as the low recall. This paper presents a new method which is based on pseudo relevance feedback by using cross media. This paper firstly analyzes the keywords in consumption intention, then formulates it as queries to submit to the search engine, finally extracts consumer target from the search results. The experimental results are significantly better than those of the baseline methods relying on existing technologies.

Key words: consumption intention, consumption target, cross social media search, information extraction, monolingual word alignment model, social media

摘要： 目前，微博消费意图识别问题成为新的研究热点。然而，已有工作主要判断微博是否具有商业意图，很少研究消费意图内容中消费对象的识别问题，而消费对象的识别是精确地进行商业推荐的关键，因此对其进行研究具有重要意义。微博内容较短且用户用语不规范，基于词典匹配与依存关系分析的方法，在识别召回率上难以得到满意的效果。提出了一种基于跨媒体伪相关反馈的方法。首先分析具有消费意图微博中的关键词，而后将其视为查询提交给搜索引擎进行搜索，最后从搜索结果中提取消费意图对象。实验结果表明，提出的方法显著地优于基准方法。

关键词: 消费意图, 消费对象, 跨社交媒体检索, 信息抽取, 单语词对齐模型, 社交媒体