User_manual.md

PyGAAP User Manual

The Python graphical authorship attribution program is a text classification tool in Python and a port of JGAAP, the original java version. It follows the typical steps seen in a text classification task: from pre-processing, feature extraction, and feature culling to text embedding and analysis.

• Please see the developer manual if you would like to extend PyGAAP.

Ten-second GUI demo

1. Launch PyGAAP,
2. Click on Files $\rightarrow$ AAAC Problems $\rightarrow$ Problem A
3. Switch to the Event drivers tab, click on Character NGrams, then Add
4. Switch to the Embeddings tab and click on Frequency, then Add
5. Switch to the Analysis Methods tab and click on Linear SVM (sklearn), then Add.
6. Finally, switch to Review & Process tab, press Process, and wait for the results!

The graphical interface

Loading document(s)

Unknown documents (the testing set)

Known documents (the training set)

Modules

Alternative names

The module types in PyGAAP have alternative names one may have seen in academic papers or other machine learning programs. Here are some of the names one may encounter.

• Canonicizers: Pre-processing; Text normalization.
• Event Drivers: Feature (set) extraction, Characteristics extraction, "Write-print" (a particular feature extraction method).
• Event Culling: Feature (set) culling/filtering.
• Embeddings: Feature embedding
• Analysis Methods: Classifiers, Algorithms.
• Distance Functions: Distances, Metrics.

Add/remove modules

Module parameters

Search modules

The command line

Corpus and experiment CSVs

Creating a corpus csv

Creating an experiment csv

parameters

Loading a JGAAP experiment csv

PyGAAP is cross-compatible with JGAAP experiment csv format. Simply run the CLI as if you'd run a PyGAAP csv. Caution that since JGAAP's exp. csv format doesn't specify an embedding, the default will always be Frequency.

Convert a JGAAP expriment csv to the PyGAAP format

Running the experiment