Grounded Text Mining Approach (GTxA)

Description

Combine grounded theory coding with dictionary-based text mining: write auto-coding rules from in vivo codes and revise them until machine and manual coding agree.

A qualitatively dominant hybrid for crossover analysis of text. It begins with a computational overview of the corpus (word frequencies, co-occurrence networks, correspondence analysis, keyword-in-context reading), then proceeds with constructivist grounded theory coding: initial line-by-line codes, focused codes and a hierarchy. Researchers then write explicit auto-coding rules from in vivo words and focused codes, apply them to the whole corpus, and quantify codes by group with cross-tabulations, chi-square tests and residuals. The machine coding is compared with manual coding and the rules are revised until they agree, which is the method's main check. Meta-inferences are drawn from the combined display and theoretical sampling continues to saturation. Tools such as KH Coder and MAXQDA support it. Its value is transparent, reproducible rules rather than opaque models; its weakness is that word-level rules lose context.

Attributes

Loss type
Siloed
Data
Mixed
Effort
Weeks

Steps

  1. Overview the corpus with word frequencies, co-occurrence networks, correspondence analysis and keyword-in-context reading.
  2. Code a substantial sample with grounded theory: initial line-by-line codes, then focused codes grouped hierarchically.
  3. Write auto-coding rules from in vivo words and focused codes and apply them to the full corpus.
  4. Quantify codes by group and cross-tabulate, testing differences with chi-square and adjusted residuals.
  5. Compare machine coding with manual coding on the same segments and revise rules until they agree.
  6. Draw meta-inferences from the combined display and continue theoretical sampling to saturation.

Sources

  • The Sage Handbook of Mixed Methods Research Design (Poth, 2023)

Also known as

  • GTxA

Categories

  • User Research
  • Qualitative Analysis & Synthesis