Processing textual (unstructured) data takes time and increases costs considerably. We mask direct identifiers in textual data, which affects how usable the data is.
Consider whether you really need textual data
Similar information can often be extracted in a structured format. You must always justify the need for unstructured data.
Ask the data controller for advice on search terms
The data controller is best placed to judge which search terms will give complete results without unnecessary information. For example, the search term “pressure” will return all entries that mention pressure, such as blood pressure and eye pressure. If you are researching blood pressure, the extracted data will then contain a lot of information you do not need.
Define the length of the text extract for each variable
Keep the extract as short as possible, for example the search term and 50 characters on either side of it. Ask the data controller what length would be suitable.
Limit the extraction
If you can limit the extraction to data from a specific department or specialty instead of the organisation’s entire database, the amount of text and the processing costs will be considerably lower.