Profiling Metadata

Currently we support automatic analysis of 5 different file types. The metadata is only populated by our automatic profiling and cannot be edited. This is because we want to preserve the objective analysis by a computer system here.

Text Files

• application/pdf
• text/plain

The automatic analyses are listed and explained below:

• A set of simple metadata is extracted from the text file (size, PDF version, creation date, ...)
• We are trying to determine the language of the text. A score is additionally stored, which indicates the confidence.
• Keywords are extracted automatically. These are displayed on the main page of the entity.
• Key phrases are extracted using two algorithms. Currently, these are LSA and Luhn.
• A simple statistical evaluation on the distribution of charcters and number of different tokens (words, sentences, and characters).
• We perform a scan for personal data and give a score between 0 and 100, where 100 means that the resource is very likely to be GDPR relevant.

Table Files

• text/csv
• application/x-sas-data

The automatic analyses are listed and explained below:

• We extract the schema of the table and visualize it.
• We perform a statistical analysis that looks at each column and extracts key metrics.
• We extract a sample set of the table to give an insight into the data.

Image Files

• image/jpeg

The automatic analyses are listed and explained below:

• We create a thumbnail so that you can see which image we are dealing with.
• We try to recognize the objects seen in the picture.
• We try to generate a caption for the image.
• We perform OCR text recognition on the image to make text on images search relevant in our catalog as well.
• We extract EXIF metadata, e.g. about the camera with which the photo was taken.