Introduction Google Maldives Shoping on taobao com Groupon
Introduction Google Maldives Shoping on taobao. com Groupon on dianping. com Hunan University Software Engineering 2
3 Current situation Adapt to English only Based on PMI & Map Accuracy is low Long run time Hunan University Software Engineering
4 Aim To adapt to Chinese To improve the accuracy To provide the reference Hunan University Software Engineering
PMI-IR Hunan University Software Engineering 5
SO-PMI Hunan University Software Engineering 6
What’s new Set of subjects 诺基亚、诺基亚手机、Nokia、NOKIA Lumia、LUMIA、撸妹、800 Set of negative adverbs 否、非、不 Factor of range Elements of Pset and Nset Public area Private area Hunan University Software Engineering 7
8 Experiment OS: linux x 64 bits CPU: 8 cores Memery: 64 GB HD: 1 TB Data: 30 GB Runtime: 7 hours Hunan University Software Engineering
9 Experiment (cont. ) 中文总词数 正面词次 负面词次 所有倾向性词次 641, 199, 184 2, 400, 000 3. 74 2, 440, 000 3. 81 4, 840, 000 7. 55 ‰ Hunan University Software Engineering ‰ ‰
11 Results 评论领域 准确率 平均SO(-1~1) 五星值(0~5) 电脑 99% 0. 344 3 数码 98% 0. 254 3 餐饮 97% -0. 187 2. 5 大家电 97% 0. 518 4 电影 90% -0. 421 2 其他 97% -0. 093 2. 5 Hunan University Software Engineering
Some problems Rhetorical Adjectives used as nouns and adverbs “可口可乐”-“可口”“可乐”“好心当作驴肝肺”-“好心 ”“…” Neuter adjs “实在-拔丝香蕉实在是难吃”“卫生-卫生条件太差” Error in posting tags “不好看吗?”“不见得能成功” “普遍”“个别”“有限”“特殊” Time and Memery Hunan University Software Engineering 12
13 Potential applications Intelligent search engine Electronic commerce The evaluation system Public opinion supervision Business intelligence analysis Hunan University Software Engineering
14 FQA Thanks all author: Shen. Tong mail: tt 0 rz@yahoo. cn site: shentong. xoom. it Hunan University Software Engineering
- Slides: 14