Quantitative Discursive Analysis (QDA)

(C) 2019 Mark M. Bailey, PhD

About

Quantitative Discursive Analysis (QDA) converts bodies of text into graph objects built from noun phrases. Each noun or modifier becomes a vertex, and edges are determined by how nouns and modifiers are linked within phrases. The more central a noun is to the overall text content, the higher its centrality measure. This makes the graph representation more robust than simple keyword frequencies.

QDA compares discursive content by calculating resonance between two texts. Resonance is the cosine similarity of the betweenness-centrality vectors for the intersection of vertices in both texts. Values are normalized to [0, 1], where 0 indicates no overlap and 1 indicates perfect overlap.

Installation

pip install .

Dependencies

• Python 3.10+
• networkx
• numpy
• textblob

Important: TextBlob corpora required for default extractor

The default noun phrase extraction method (textblob) requires TextBlob/NLTK corpora.

python -m textblob.download_corpora

If corpora are unavailable, use the fallback extractor documented below.

Quickstart

Default extractor (textblob)

import QDA

text_a = "This is a string of text about politics and economics."
text_b = "This is a different string of text about music and art."

g1 = QDA.discursive_object(text_a)  # noun_extractor='textblob' by default
g2 = QDA.discursive_object(text_b)

print(QDA.resonate(g1, g2))

Fallback extractor (simple, no corpora required)

import QDA

text_a = "This is a string of text about politics and economics."
text_b = "This is a different string of text about music and art."

g1 = QDA.discursive_object(text_a, noun_extractor="simple")
g2 = QDA.discursive_object(text_b, noun_extractor="simple")

print(QDA.resonate(g1, g2))

API summary

• QDA.discursive_object(text, noun_extractor="textblob")
• QDA.resonate(g1, g2)
• QDA.resonate_as_series(G_list)
• QDA.resonate_as_matrix(G_list)
• QDA.discursive_community(G_list)

Development

Run tests with:

pytest

Notes and limitations

• Large texts may be slow because betweenness centrality is computationally expensive.
• Results depend on noun phrase extraction method (textblob vs simple).
• simple is a compatibility fallback and may produce different phrase quality than TextBlob.

Changelog

• 0.1.0
  • Added Python 3.10+ packaging support and pytest test suite.
  • Added explicit TextBlob missing-corpora error messaging and optional simple extractor.
  • Added NetworkX compatibility shim for NumPy graph conversion.
  • Improved performance of matrix resonance and graph construction helpers.
  • Added GitHub Actions CI for Python 3.10/3.11/3.12.