Challenges of the Email Domain for Text Classification

Abstract

Interactive classification of email into a user defined hierarchy of folders is a natural domain for application of text classification methods. This domain presents several challenges. First, the user’s changing mail filing habits mandate classification technology adapt in a dynamic environment. Second, the classification technology needs to be able to handle heterogeneity in folder content and folder size. Performance when there are only a small number of messages in a folder is especially important. Third, methods must meet the processing and memory requirements of a software implementation. We study three promising methods and present an analysis of their behavior with respect to these domain-specifc challenges.

Cite

Text

Brutlag and Meek. "Challenges of the Email Domain for Text Classification." International Conference on Machine Learning, 2000.

Markdown

[Brutlag and Meek. "Challenges of the Email Domain for Text Classification." International Conference on Machine Learning, 2000.](https://mlanthology.org/icml/2000/brutlag2000icml-challenges/)

BibTeX

@inproceedings{brutlag2000icml-challenges,
  title     = {{Challenges of the Email Domain for Text Classification}},
  author    = {Brutlag, Jake D. and Meek, Christopher},
  booktitle = {International Conference on Machine Learning},
  year      = {2000},
  pages     = {103-110},
  url       = {https://mlanthology.org/icml/2000/brutlag2000icml-challenges/}
}