README.md

metadata

license: apache-2.0

Inverted Index Dataset

This dataset contains the inverted index part of the manuscript.
The original data was stored as JSON files and has been compressed into a ZIP archive for upload and distribution.

The inverted index maps words (obtained by tokenizing chart descriptions) to the images in which they appear and the frequency of occurrence in each image.

---

Data Format

Each JSON file stores an inverted index with the following logical structure:

• Word
  • Specific image location (relative image path)
    • Frequency (number of times the word appears in that image)

Conceptually, the structure can be represented as:

{
  "<word_1>": {
    "<image_location_1>": <frequency>,
    "<image_location_2>": <frequency>
  },
  "<word_2>": {
    "<image_location_3>": <frequency>
  }
}