Email Data Cleaning Tsinghua University电子邮件数据清理清华大学.pptVIP

* On the J2EE dataset, we see that about +24.7% improvements can be obtained in terms of F1-Measure. The experimental results are consistent with the results on the BR dataset. * Finally, we got a summary for our work. * In this paper. We have investigated the problem of email data cleaning and formalized it as non-text filtering and text normalization. We have proposed a cascaded approach to the task. Using Support Vector Machines, we have been able to make an implementation of the approach. Experimental results show that our approach can significantly outperform baseline methods for email data cleaning. When applying it to term extraction from emails, we observe a significant improvement on extraction accuracy. * * Another interesting work is to make the detection adaptive to new domains. We analyzed headers in different datasets. We have found that the header of different datasets have different patterns. Header models learned from one dataset may work not well on the other datasets. We have conducted a preliminary experiment. We collected a new dataset that is called WWW. The WWW dataset is from enterprise track in TREC 2005. We applied the trained SVM models directly to the WWW dataset. We found that the drop is about 10% in terms of F1-Measure. We plan to deal with the problem by making use of semi-supervised methods. Outline Motivation and Problem Description Related Work Our Approach Implementation Experimental Results Summary Summary Formalized email data cleaning as non-text filtering and text normalization Conducted email cleaning in ‘cascaded’ approach Used SVM models for header, signature, program code, and extra line break detection Our approach significantly outperforms baseline methods When applied to term extraction, significant improvement on extraction accuracy can be obtained Thanks! Examples of List and Table in Email Supppoesdly this is what it does: *New Layout *Pentium-safe Compiled *Fixed Various bugs in ftp scan *Fixed R