Introduction

This web application is designed for producing interactive visualisations of word co-occurrence networks derived from ancient Greek texts. It is based upon a corpus of 687 preprocessed texts, publicly available via GitHub from here.

The word co-occurrence networks are generated and visualised by means of Python functions developed by the DiRECT project, and distributed as an independent and free-to-use Python package called TextNet (GitHub, test.pypi) . The functions rely upon many other Python packages, especially networkx (for working with network data) and Plotly (for interactive visualisations). The web application is based on Pythonanywhere.

Word-Co-Occurrence Networks and N-Nearest Neighbors

A word co-occurrence network based upon a document is typically too large to be meaningfully visualised as a whole. Therefore, it makes a lot of sense to plot only a sub-selection of it. With this in mind, this application is designed to plot only a subnetwork consisting from three types of elements:

• key term chosen by the app user
• a set of its nearest neighbours within the network
• a set of edges representing shortest path the key term and its nearest neighbours

There is an idea that such a sub-network can say us something about meaning of the key term within given document, which can be compared across documents.

App tutorial

To use this app, you have to configure a set of parameters: select an author, her/his work, particular key term to focus on and a number of the key term’s neighbours. Afterwards, click on the button “Generate graph”.