Matches in ScholarlyData for { <https://w3id.org/scholarlydata/inproceedings/www2008/paper/765> ?p ?o. }
Showing items 1 to 16 of
16
with 100 items per page.
- 765 creator gui-rong-xue.
- 765 creator qiang-yang.
- 765 creator wenyuan-dai.
- 765 creator xiao-ling.
- 765 creator yong-yu.
- 765 creator yun-jiang.
- 765 type InProceedings.
- 765 label "Can chinese web pages be classified with english data source?".
- 765 sameAs 765.
- 765 abstract "In the China Web Mining, classification often meets the situation of lacking labeled data, while there are plenty of labeled English Web pages available on the Web, e.g. the Open Directory Project. The information contained in the Chinese and English Web pages may differ somewhat, while their feature representations are different as well. In this paper, we propose a transfer learning method to solve this cross-language classification problem. The English labeled data are firstly translated to the low quality Chinese data, and then domain-transfer learning is applied to transferring only useful and accurate knowledge from translated English Web pages to the Chinese ones. Compare with previous semi-supervised learning methods, our method better considers the domain differences and transfers only useful knowledge and removes noises. The theoretical and empirical studies shows that our method is effective. Compared with several state-of-art algorithms, our approach shows significant improvements, while scaling very well.".
- 765 hasAuthorList authorList.
- 765 hasTopic World_Wide_Web.
- 765 isPartOf proceedings.
- 765 keyword "Cross-Language Classification".
- 765 keyword "Information Bottleneck".
- 765 title "Can chinese web pages be classified with english data source?".