marty-oehme/pubs-extract

Extract annotations from any pdf document. A plugin for pubs bibliography manager. Mirror of https://git.martyoeh.me/Marty/pubs-extract

pubs-extract's Issues

Improve explanation of note/pdf-embedded text similarity

Especially the minimum_text_similarity configuration should be improved, if somebody does not understand the way that pdf annotations have both a 'highlighted' text and an appended 'note' text, they will be lost with similarity explanations and the different possible outcomes.

Perhaps a footnote and a little picture to help visualize it or something similar would suffice.

Add some basic tests to output formatting and annotation recognition

Currently, we do not have any tests. Before expanding the scope of the project much further, it would be prudent to capture current functionality in some tests to make it easier to make refactors, see changes and find problem areas.

Allow page-sequential annotation extraction in notes

Currently, new annotations always get appended to the end of notes. They do so in the order that they appear in the PDF document (i.e. sequentially).

This option could allow adding annotations in a sequential fashion for the whole note, i.e. a new annotation on page 15 would get situated between the existing annotations for page 13 and page 18.

Requires some way of clearly delineating what is an annotation and what is manually added in a note.

Allow custom colors for tags

Currently, we provide 6 colors that annotations can be recognized in:
red green blue yellow purple orange.

Since we are already mapping these color names to exact rgb tuples behind the scenes,
we may as well allow the user access to provide exact tuples to check for custom annotation colors.

A possible configuration file could then look like this:

[[[tags]]]
orange = "important"
blue = "todo"
mygreycolor = "0.5,0.5,0.5:unimportant"

in the form {customcolorname} = "{color_vector}:{tag_name}. This form would preserve the old settings structure but be a little awkward with no purpose for the color name given.

A more streamlined form would be:

[[[tags]]]
important = "orange"
todo = "blue"
unimportant = "0.5,0.5,0.5"

which turns tags and mapped color around, making it easily possible to map custom colors but breaking the old configuration format.

Improve annotation duplication detection in notes

Currently, the plugin does a 1:1 exact match for existing annotations in notes.

Since we already have Levenshtein distance calculation we can potentially use this to get a more lenient annotation comparison (e.g. fixed spelling mistakes, or added words missing in auto-extraction).

The hardest part would presumably be to find the range which should be compared to the extracted note - we can not compare the whole document, so which existing annotations should we compare and how do we delimit annotations?