By Aldo Benini
“The purpose of this paper is to add simple productivity tools for text analysis, by publicizing existing ones and by adding one that I created. “Simple” is a relative term. As the diagram in the Summary section suggests, the suitability of the tools depends on the skills and equipment level of the intending user. Also, I assume a kind of working environment that developing country organizations will not everywhere offer for computer-supported text analysis: that the analyst actually can acquire the documents digitally.
The need for these tools arises in two distinct situations:
• The analyst is comfortable with classic interpretive (i.e., non-statistical)approaches, focused on the intent of the document authors, the texts’ internal logic, and the relevant professional and audience contexts. The need is for rapid navigation and comparison inside and between texts. For a typical situation, think of the briefing papers that an evaluation team receives, some with the terms of reference, some at the project site.
• Second, purely interpretive approaches break down, because of the volume or structure of the text material, or are not chosen in the first place. The analyst, with the help of statistical tools, investigates distributions and correlations of text elements that carry meaning, notably words and terms. Use of such results may range from navigation to exploration to testing of hypothesis. For example, a federated international NGO may produce major policy documents centrally.
Over time, local policy adaptation and implementation are reported in numerous documents created in the participating member organizations. Their volume defies direct interpretive access.
The tools presented here serve both the strictly interpretive analyst and the statistically minded, in differing degrees and mixtures. In fact, they should help building bridges between methodological communities. The reader may choose which of the three tools, if any, will possibly benefit his work. These three are located at different heights of a learning curve. The easier ones work equally well regardless of whether the more demanding ones are adopted or not. The more demanding ones demand skills in spreadsheet and statistical applications.”
