A reading instrument

Close reading, at corpus scale.

Does the document talk about a topic — and does it talk about it in the right place? Document Lens turns a folder of PDFs, a keyword list, and the axes you care about into coverage heatmaps, scores, trend lines, and concordances. Built for researchers studying corporate disclosure; at home in any keyword-driven study of unstructured text.

Local-first. Your documents and every analysis live in SQLite on this device.

The research loop

01

Import

Drop in a folder of PDFs, Word files, or slides. Text is extracted per page and the embedded images come with it, anchored to the page they appeared on. A shared library means each document is processed once, reused across every project.

02

Frame the question

Bring your keyword list, the axes you want to break the analysis along, and a scoring rule. Polarity separates terms that signal delivery from counter-terms that signal performative language — one list, both narratives.

03

Focus

The front door. Every document ranked by notability, so you start with the ones worth reading instead of the ones that happen to be first alphabetically.

04

Drill down

Every finding is a link. A signal opens the comparison behind it, a document title opens the concordance. Read the evidence, adjust the frame, re-rank — then export the tables and figures into your paper.

Twelve workflows, one hub

Measure

Coverage heatmaps for every keyword × document pair. A single Wedding Cake score per report when you need one number, year-over-year tracking when you need the trend, and rankings when you need the comparison.

Interrogate

Audit whether each keyword is used in the right context — anomalies and confirmations side by side. Ask where the tone runs ahead of the substance. Then open the concordance and read what the document actually says, with the PDF right alongside.

Map

Cross two axes — SDG × Function, Pillar × section, any pair you define — and see how each document distributes its attention. Talking about a topic is one thing; talking about it in the right place is the finding.

Local-first, by design

Research corpora are often confidential, embargoed, or simply yours. Document Lens is built so your documents and your analysis never leave your machine.

For disclosure researchers

The original use case: sustainability reporting in university annual reports. Keyword polarity supports both narratives from one list — terms that signal delivery, and counter-terms that signal greenwashing — with the Wedding Cake score grounding it all in the SDG model.

For any keyword-driven study

Policy documents, curricula, submissions, transcripts — if you have a corpus, a framework, and a question about who says what, where, the same projects, axes, and scoring rules apply. Bring your own taxonomy.