stemmabench 0.0.1

Package for benching stemmatology algorithms

StemmaBench

StemmaBench is a Python package for quick generation of artificial, synthetic scribal traditions. From an original text, it generates a series of witnesses based on a set of variable aiming at describing the different scribal behaviors.

Motivation and theory

Quickstart

Install

Install can be done through the invoke task manager:

You first need to install invoke. pip install invoke

Then run the command install invoke install (add --venv to run within a virtualenv)

If you want to run the tests, install the package in test mode: invoke install --extra test

You can run the tests: invoke test (add --venv to run within the same virtualenv)

Command line

A possible way to generate a tradition is running the generate command:

generate --help
Usage: generate [OPTIONS] INPUT_TEXT OUTPUT_FOLDER CONFIGURATION

  Generate a tradition of manuscripts.

  Args:
  input_text (str): The text to give as input for the tradition.

  output_folder (str): The output folder for the tradition.

  configuration (str): The configuration of the tradition.

Arguments:
  INPUT_TEXT     [required]
  OUTPUT_FOLDER  [required]
  CONFIGURATION  [required]

For demonstration, go to the folder demo and run: generate .\test_text.txt output_folder .\config.yaml

Interactive use

See: jupyter quickstart

To do list

[] Add support for ancient languages (greek, hebrew mainly) => Will be done through using MACULA API

[] Add the possibility to work with contamination: [] Model contamination [] Create new contaminated text

[] Add support for missing manuscripts (new variable in stemma generation that will skip a generations): we do not have access to some heuristics.

[] Make vizualization more resiliant + a whole part of the package