Stylometry Project

1
Stylometry Project

• IT691 CS615Computer Information Systems
  Projects
• December, 2007

2
Team Members

• Geraldine McCabe Team Leader
• Huriya Manzar Programmer
• Melissa Connors Programmer
• Kristina Calix Implementer
• De Havaland Levy Quality Assurance

3
Overview

• This program can be used by any researcher
  attempting to identify the authorship of email
  text messages

4
Overview of Existing Program

• A pattern recognition system to identify the
  author of arbitrary email using Stylometry
  features
• Existing C program used raw keystroke data and
  converted into simple text files
• Performs feature extraction for statistical
  analysis, followed by classification using
  K-nearest neighbor

5
Program Modifications

• Collected larger data set of plain text email
  samples for improved accuracy of testing, 10
  samples from each of 12 different authors
  averaging 150 words
• Keystroke features were removed from existing
  program and new features added to provide a total
  of 55 stylistic features for extraction.

6
Modifications Cont.

• Demographics for each author was added as per
  clients request
• Reset option was added to allow for single input
  of demo info for multiple samples from each
  individual author.
• Feature vector data was normalized in the range
  0-1 and formatted to provide a CSV file.

7
Modifications Cont.

• GUI was enhanced to eliminate unnecessary menu
  options provide relevant options for new
  modifications

8
Demonstration

• Plaintext email samples
• Create Base Data Set
• Normalize Base Data Set
• Output normalized data as CSV/Excel file
• Compare unknown author

9
Future Work

• Add additional features for per clients requests
  
• Since formatting plays a big part in Stylometry.
  features such as indentations, number of blank
  lines between paragraphs, number of blank lines
  between the last sentence and the closing, number
  of spaces after periods (some people type 1
  space, some people type 2 spaces), could be
  added
• Grammatical features For example, stylometry
  experts have noticed that women tend to use
  adverbs more than men.
• Identify gender based on stylistic linguistic
  habits

10
Questions

• Contact gm60518w_at_pace.edu for more information or
  visit http//utopia.csis.pace.edu/cs691/2007-2008/
  team2/index2.htm