The output .json file can contain multiple matches. This tool itself cannot determine which one is correct, but it incorporates scoring facilities through the following features:

• The score option of the matching rule, or the sum of scores if multiple rules match in a compound rule
• Fuzzy errors
• The positional rule's proximity to the anchor
• The top and left properties of the bounding box
• Anchor's features, as mentioned above

Considering the above, matches are finally sorted in the following order:

1. Score
2. Error (derived from the voc rule)
3. Proximity
4. Anchor score
5. Anchor error
6. Anchor proximity (of a parent anchor, if any)
7. Number of matches (provided by a fuzzy match)
8. Size of the rule list
9. Top
10. Left

This sorting mechanism even applies in Blocks, expression, where the use of the same-block positional rule might suggest something different in practice, such as in the case of addresses. This is why we provided a command line with jq in that example.