Results for 'Email Formality in the Workplace: A Case Study on the Enron Corpus' URL for paper http://aclweb.org/anthology-new/W/W11/W11-0711.pdf
Kelly Peterson [email protected]
Matt Hohensee [email protected]
Fei Xia [email protected]
Data references : ISI database : Retrieved December, 2010 from:
- http://www.isi.edu/˜adibi/Enron/Enron.htm University of Sheffield personal vs business annotations : Retrieved December, 2010 from:
- http://staffwww.dcs.shef.ac.uk/people/L.Guthrie/nlp/research.htm
File Contents:
/README
Dearest reader, you are reading me as we speak
/annotations/
All of the human annotated files for formality and requests
NOTE : when an email ends in .txt, this is numbered by the 'mid' column in the ISI database.
Otherwise, the filenames are the same as the originals from the CMU dataset
/formality/
/100_emails_agreement
Initially, 3 annotators ennotated 100 emails for formality annotation agreement
/400_emails_training
After agreement, one annotator had time to annotate another 300 emails totalling 400 so that the classifier could be trained
/requests/
2 annotators annotated for the presence of a request.
In these files, a '1' right of the file name indicates a request, otherwise there was no request
/furcoat
Python scripts for generating the feature vectors for the formality classifier.
/mysql/
/queries/
/get_formality_and_requests_by_position.sql
This query was used to derive the results in Table 6 of the paper
/get_formality_by_rank_diff.sql
This query was used to derive the results in Table 7 of the paper
/requests_and_formality.sql
This query was used to derive the results in Table 8 of the paper
/get_recipient_count_and_formality.sql
This query was used to derive the results in Table 9 of the paper
/tables/
3 tables were added to the ISI database during our case study. All of these can be used to JOIN against MySQL tables provided by ISI.
No indexing is added to any of the columns in these .sql so please note that queries will run EXTREMELY slowly until indexing is added.
You will likely want to add an INDEX to the following columns : 'mid', 'Address' and 'Rank'
/enron_formality.sql
This table comprises the formality classification results in the column 'EffectiveLabel' (0=Empty, 1=Formal, 2=Informal)
It also contains counts of the various features that the classifier extracted
/enron_positions.sql
A table created based on the positions in the ISI spreadsheet
/enron_requests.sql
This table comprises the requests classification results in the column 'RawLabel' (0=NonRequest, 1=Request)
/python/
For more information on running these scripts or reproducing these results, please contact Kelly Peterson
/contact_frequency/
This script was used in combination with a CSV file exported from MySQL to derive the results in Table 5 of the paper
/personal_vs_business_formality/
This script was used in conjunction with Mallet's output from running against the
University of Sheffield Personal vs. Business dataset to derive the results
/ scripts
Lots of random scripts for Wordnik informal word API scraping, preprocessing, data reporting, etc.
NOTE : It's been a LOOOOONG time since looking at these so please use at your own risk!
/enron_employee_positions/
Included here are positions and ranks from Diesner et al and some re-ranked positions that we performed in our work
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent