The Enron Corpus: A New Dataset for Email Classi(cid:12)cation Research

By Bryan Klimt, Yiming Yang. 2004

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

Citations in the corpus

PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization (2019, background): “AESLC (Zhang and Tetreault, 2019) consists of 18k email bodies and their subjects from the Enron corpus (Klimt and Yang, 2004), a collection of email messages of employees in the Enron Corporation.”
Datasheet for the Pile (2022, background): “Enron Emails: The Enron Emails dataset [Klimt and Yang, 2004] contains text from the contents of Enron’s email servers unearthed during the investigation into that organization’s accounting methods and is a valuable corpus for understanding the modality of email communications, which are typically…”
QUINTA: A question tagging assistant to improve the answering ratio in electronic forums (2015, methodology): “In an IR system the vector for each document is compared with the one generated by a query, usually through a common technique known as TF/IDF (Term Frequency/Inverse Document Frequency, [8]) and a similarity metric such as the cosine distance [9].”
Feature Cross-Substitution in Adversarial Classification (2014, methodology): “The Enron data set was divided into training set of 3172 and a test set of 2000 emails in each of 5 folds of cross-validation, with an equal number of spam and non-spam instances [21].”
Simplicial closure and higher-order link prediction (2018, background): “• Email networks [email-Enron (23); email-Eu (24)]: Nodes are email addresses and a simplex is a set consisting of all recipient addresses on an email along with the sender’s address; emailEnron spans most of the duration of a company’s lifetime, and email-Eu spans over 2 years.”
Multi-step Jailbreaking Privacy Attacks on ChatGPT (2023, methodology): “Tables 1 and 3 depict the email address recovery results on the filtered Enron Email Dataset and manually collected faculty information of various universities.”
Spam Filtering Based On The Analysis Of Text Information Embedded Into Images (2006, methodology): “As suggested in Klimt & Yang (2004), to avoid many duplicate e-mails we first discarded from the Enron data set the ones in the “deleted items”, “all documents” and “sent” folders, reducing it to about 200,000 e-mails.”
Addressing imbalance in multilabel classification: Measures and random resampling algorithms (2015, unknown): “9596 [29]”
Interacting meaningfully with machine learning systems: Three experiments (2009, methodology): “122 messages from a user’s email (farmer-d), which had sufficient content for human and machine classification, were drawn from the publicly available Enron Page 5 dataset (Klimt and Yang 2004).”
Tracking recurring contexts using ensemble classifiers: an application to email filtering (2009, methodology): “Since BL can operate without an incremental algorithm we additionally exploited support vector machines (SVM) [38] which are considered more accurate classifiers in text classification [39,40] and also used in email classification [ 41 ].”
Reusable workflows for gender prediction (2018, background): “Enron-mails dataset (Klimt and Yang, 2004)5 contains 517,431 e-mails of Enron employees, 105 male and 43 female.”
The structure of information pathways in a social communication network (2008, methodology): “We will use this university e-mail dataset as the primary focus of discussion in the present work; but at the end, we also discuss the results of our analyses on two other sources of data: the Enron e-mail corpus [16], a widely-used dataset containing of e-mail communication among executives from th”
Communication Networks from the Enron Email Corpus “It's Always About the People. Enron is no Different” (2005, background): “Separately, Shetty and Adibi (a) and Klimt and Yang (2004a) cleaned the corpus, explored the distribution of messages across users and time, and characterized discussion threads (Klimt and Yang, 2004b).”