Remove fake subjects from Works

[tfmorris]

Description

Three of the top five "subjects" are not subjects at all:

• Accessible book - 2.5 million
• Protected DAISY - 1.2 million
• In library - 0.5 million

The are also some lower frequency noise terms like Lending library and Internet Archive Wishlist but the three above represent the bulk of the noise.

Expectation

The subject list should contain things which are actually subjects of the work.

Proposal & Constraints

Remove the three subjects above. If they're needed to provide functionality move them to a hidden portion of the Solr index where they don't pollute the UI.

[LeadSongDog]

@tfmoorris Perhaps these tags should be supplanted by a new entry under The Physical Object / Format prior to deletion, or is there a better way?

[hornc]

We should also remove this code which looks like it is trying to protect these subjects:

openlibrary/openlibrary/plugins/upstream/addbook.py

Line 38 in 47c72f9

SYSTEM_SUBJECTS = ["Accessible Book", "Lending Library", "In Library", "Protected DAISY"]

openlibrary/openlibrary/plugins/upstream/addbook.py

Line 597 in 47c72f9

self._prevent_system_subjects_deletion(work)

openlibrary/openlibrary/plugins/upstream/addbook.py

Line 600 in 47c72f9

def _prevent_system_subjects_deletion(self, work):

These subjects should no longer be used as there are other ways to get the information they were trying to convey.

[LeadSongDog]

@hornc
Are those "other ways to get the information" all working and documented to users?

[hornc]

Checking ol_dump_editions_2019-12-31.txt

• Lending library appears to be clear
• Internet Archive Wishlist 0, also clear

ol_dump_works_2019-12-31.txt

• Lending library 0
• Internet Archive Wishlist - 382004 remaining
• In Library - 436321 remaining

[hornc]

Subjects search for quicker testing of current numbers:

• https://openlibrary.org/subjects/accessible_book
• https://openlibrary.org/subjects/protected_daisy
• https://openlibrary.org/subjects/in_library

NB the solr indexing of these subjects seems way off -- there are many more items showing up with these subjects at the URLs than are in the data dumps.

[hornc]

The last accessible book subjects found in the ol_dump_works_2020-07-31.txt.gz have been processed and had these subjects removed.