What corpus is and why it exists
Which research tasks the platform covers, which programs it replaces, how it handles data and where the data is stored.
corpus is a web platform for running an empirical sociological study. It brings together tools that are usually spread across separate programs: a register of participants and consent, interviews and transcripts, field notes, surveys, network data, coding, quotes, a bibliography and data export for statistical analysis. Below: the problem it solves and how it works.
The problem: one study, many programs
A typical study with interviews and a survey looks like this. Recordings sit on a voice recorder or in a cloud service. Transcripts are in a word processor, often in several versions. Coding happens in qualitative analysis software. The survey runs in an online survey tool, and responses are cleaned by hand in a spreadsheet. Networks are drawn in separate graph software, references live in a reference manager, and signed consent forms sit in a paper binder.
Each program does its own job well, but they are not connected. What links them is file names and the researcher’s memory. This creates practical problems:
- the same participant appears in different files under different labels, and matching their interview, survey answers and consent form is manual work;
- nobody knows which version of a transcript or spreadsheet is final, or what was changed in it;
- a few months later it is hard to say which questionnaires a percentage in a chapter was computed from, or where a quote came from;
- participant names and contact details end up in file names, emails and personal devices.
The solution: one platform for the whole study
In corpus, a study is a single workspace. It contains:
- the research path: topic, problem statement, research question, hypotheses and methodology; changes to the methodology are saved as versions with a reason;
- participants and consent: each participant has a code (for example P-014); names and contact details are stored separately and encrypted;
- interviews and transcripts, versioned: raw, cleaned, anonymised, translated;
- field notes, with sections for observation, quotes, interpretation and reflexivity, plus a field archive;
- surveys: a builder, collection by link, entry of paper questionnaires, cleaning rules and export to R;
- coding and quotes: a codebook, coding of passages, codes × texts matrices, inter-coder agreement;
- networks: an “individual — tie type — individual 2” table becomes a graph with its measures;
- bibliography: RIS and CSL-JSON import from reference managers, ISO 690 or APA 7 formatting.
All materials are linked through the participant code. A participant’s page opens their interviews, survey answers, quotes and consent.
How data is handled
corpus takes its data rules from clinical trials, where they have long been fixed in the ALCOA+ standard (see the dedicated article). In practice:
- Raw data is never overwritten. The uploaded file, the first transcript version and the raw survey answers stay unchanged.
- Every processing step creates a new version. Cleaning a text or a survey does not edit the data in place; it creates a version that records what it was made from and how.
- Every action is logged: who, what, when, the value before and after. Entries are linked in a hash chain, so a past entry cannot be changed unnoticed.
- Decisions are recorded with their reasons. For example: “we exclude questionnaires completed in under a minute — a sign of careless answering”.
- Consent applies to every export. If a participant withdraws consent, their material is removed from exports, quotes, statistics and the API.
Where the data is stored
The platform and study data are hosted by a French provider on servers in France and are not transferred outside the European Union. The university or lab remains the data controller under the GDPR; corpus acts as the processor, and the corresponding agreement is being prepared for the pilot. Sign-in uses a password and a one-time code sent by email.
What corpus is not
corpus does not analyse interviews for the researcher: interpretation remains their work. It does not replace statistical packages, programming languages or a word processor: data exports to open formats — CSV, JSON, DOCX, GraphML, an R script — and stays readable without corpus.
Why the name
Corpus is Latin for “body”. In sociology and linguistics the word means the collection of material an analysis is built on. That is exactly what the platform keeps: everything a study’s conclusions are made of.
How to try it
corpus is currently preparing a pilot with research teams. Access is by invitation. Write to the address on the contact page: we will open a demo study with fictional data and, for the pilot, a dedicated space for your lab.