Add plain citation parsing with LLM by InAnYan · Pull Request #11831 · JabRef/jabref

InAnYan · 2024-09-25T19:12:13Z

Closes #11825
Closes #11805
Implements #11825 (comment).

Updated dialog for online parsing:

Updated dialog for offline parsing:

New setting:

Mandatory checks

Change in CHANGELOG.md described in a way that is understandable for the average user (if applicable)
~~- [ ] Tests created for changes (if applicable)~~
Manually tested changed features in running JabRef (always required)
Screenshots added in PR description (for UI changes)
Checked developer's documentation: Is the information available and up to date? If not, I outlined it in this pull request.
Checked documentation: Is the information available and up to date? If not, I created an issue at https://github.com/JabRef/user-documentation/issues or, even better, I submitted a pull request to the documentation repository.

…mSearchBasedFetcher

koppor · 2024-09-25T20:09:20Z

"Parser choice" is nice.

always display that dropdown.
Remove "offline" / "online" in title
Add "offline" to dropdown
Adapt text ("Parser choice")
Add "Grobid" to Parser choice if enabled (GROBID enable din preference)
Add "LLM" to Parser choice if enabled (AI enabled in preferences)
Remove Dialog asking to enable Grobid

Siedlerchr · 2024-09-25T21:08:38Z

I would propose a tab layout (e.g. like in OO style selection dialog) between Grobid/OpenAI

koppor · 2024-09-26T05:49:28Z

I would propose a tab layout (e.g. like in OO style selection dialog) between Grobid/OpenAI

The selection of the service is done very raRely. Thus. I Like the approach. The selection should persist across sessions.

ThiloteE · 2024-09-26T21:27:49Z

Haven't tested the PR yet, but the screenshots show a divergence between the header of the window and the warning.

"Plain references parser (online)" vs "plain citation parsing"

It should not differ.

ThiloteE · 2024-09-26T21:35:38Z

Also, I am not sure, if this preference should be in "Web search"
And if we use LLM's it can also be "online", right?

koppor · 2024-09-27T06:31:13Z

Also, I am not sure, if this preference should be in "Web search"

What do you mean by "this"?

Current state:

User can enable Grobid
Grobid preference is in "Web search"
User can enable AI
This functionality makes use of Grobid and/or AI.
AI is used at other places.
Grobid functionality is also used at other places (e.g., importing a PDF into JabRef)

* And if we use LLM's it can also be "online", right?

Grobid is also online.

Therefore, I proposed to remove both "online" and "offline" and move that decision to the dropdown.

ThiloteE · 2024-09-27T19:59:32Z

With "this" I meant the new preference that was showing in #11831 (comment) under web-search.

ThiloteE · 2024-09-27T19:59:55Z

AI (LLMs) can be both online and offline.

koppor · 2024-09-29T12:31:19Z

AI (LLMs) can be both online and offline.

Therefore "online" and "offline" are requested to be removed. See #11831 (comment)

koppor · 2024-09-29T12:33:34Z

With "this" I meant the new preference that was showing in #11831 (comment) under web-search.

Since we don't have a tab "Import" and I don't think, this is so important, I propose it too keep it in the tab. However (!), please move it to the "Remote services" part. There, we already have Grobid-related stuff.

InAnYan · 2024-10-03T07:46:12Z

This is not GUI.

I also propose to change the name to RuleBasedBibtexExtractor.

This aligns very well with the history of natural lanuguage processing (NLP):

Rule-based methods (old BibtexExtractor).
Statistical methods (like GROBID, however this is not right, they can also use neural networks).
Neural-network methods (like LLM).

InAnYan · 2024-10-03T08:00:36Z

I also propose to change all names to "Plain Citation Parsing". Synonyms are:

"BibTex extractor" - it's too general
"Extract BibTex from plain text" - "plain text" - really? I think you meant a citation, not just an arbitrary text.
"Plain references parser" - another name. This is good actually, because it correlates to "References" section in papers. But I still propose to name it "citations", because that section could also be named "Bibliography". Citation is a very specific term and it depicts better what JabRef does.

InAnYan · 2024-10-03T08:08:59Z

WAIT

All of this time offline parsing did not support parsing of several entries (separated by two blank lines)? WTF

InAnYan · 2024-10-03T08:15:48Z

WTF? "JabRef server" - I thought it connected to some kind of GROBID server

UPDATE: Yes, it really connects to JabRef server. But I still renamed that to be more general

InAnYan · 2024-10-03T08:29:11Z

 Add "Grobid" to Parser choice if enabled (GROBID enable din preference)
Add "LLM" to Parser choice if enabled (AI enabled in preferences)

Ah, this is not that easy

Siedlerchr · 2024-10-03T08:55:53Z

We had an instance of grobid running on azure but currently I cannot deploy it again due to some weird errors from docker

…dd-llm-citation-parsing # Conflicts: # src/test/java/org/jabref/logic/importer/WebFetchersTest.java

InAnYan · 2024-10-03T09:21:30Z

Did some crazy refactoring.

Hope you like it

koppor

Some small nitpicks.

And

When opening the dialog, it should check whether the current value is still available. Example: GROBID enabled, dialog opened, GROBID selected, then preferences: GROBID disabled. Opening dialog again: GROBID should not be selected. (Note that GROBID is not in the dropdown, which is OK. However, it is selected)
- I think, there should be a preference disabled-dependency: . GROBID disabled and default GROBID: try AI. If AI not enabled, set default to rule-based. - similar to AI: If AI get disabled -> if default AI then check if GROBID enabled. Then set GROBID - else set local.
Link org.jabref.logic.importer.plaincitation.RuleBasedPlainCitationParser and org.jabref.logic.importer.fileformat.BibliographyFromPdfImporter#parseReference with a TODO (at both sides, with {@link}. They are semantically equal, but the code is different. Someone should unify the functionality.
- Similar: Link org.jabref.logic.importer.plaincitation.SeveralPlainCitationParser and org.jabref.logic.importer.fileformat.BibliographyFromPdfImporter#getIntermediateData

Note: for reviewing this PR, refactoring miner is better.

E.g., instead of thinking a file was deleted, there is detection of major refactorings

InAnYan · 2024-10-06T10:43:59Z

@koppor, I thought it will be easy to filter out parser choices

But we can't (currently, or without a hack) link preferences from different tabs.

At first, I used preferences.getAiPreferences().useAiProperty().addListener(...) (or not useAi, some other name), but when we edit preferences, we don't set them in CliPreferences...

InAnYan · 2024-10-06T10:44:46Z

Okay, GROBID is on the same page, but AI is not.

Is a hack worth it?

koppor · 2024-10-06T11:15:16Z

Okay, GROBID is on the same page, but AI is not.

Is a hack worth it?

I don't fully understand ^^. - I just guess: It is OK for me if the dropdown for default in the web search contains an entry which is available at another setting...

koppor · 2024-10-06T11:17:23Z

@koppor, I thought it will be easy to filter out parser choices
But we can't (currently, or without a hack) link preferences from different tabs.

It should be possible to get the AIPreferences passed.

At first, I used preferences.getAiPreferences().useAiProperty().addListener(...)

Sounds good!

(or not useAi, some other name), but when we edit preferences, we don't set them in CliPreferences...

I don't understand. Its a hiearchy - and the class for GuiPreferences is a superset of the CliPreferences

koppor

Very small improvement. I think, this is NOT a dirty hack, but a good solution.

…Model.java Co-authored-by: Oliver Kopp <kopp.dev@gmail.com>

JabRef/jabref#11831 Move And Rename Method public performSearch(searchQuery String) : List<BibEntry> from class org.jabref.logic.importer.fetcher.GrobidCitationFetcher to public parseSeveralPlainCitations(text String) : List<BibEntry> from class org.jabref.logic.importer.plaincitation.SeveralPlainCitationParser Extract Method public splitCitations(text String) : Stream<String> extracted from public performSearch(searchQuery String) : List<BibEntry> in class org.jabref.logic.importer.plaincitation.SeveralPlainCitationParser Rename Parameter searchQuery : String to text : String in method public parseSeveralPlainCitations(text String) : List<BibEntry> from class org.jabref.logic.importer.plaincitation.SeveralPlainCitationParser

InAnYan and others added 6 commits September 25, 2024 21:37

Add plain citation parsing with LLM

7d0c04f

Final fixes

05f2fd4

Fix checkstyle

209f7e9

Use sorted set for getSearchBasedFetchersReturnsAllFetcherDerivingFro…

75d368f

…mSearchBasedFetcher

Adapt test

832ee8a

Update CHANGELOG.md

d6ee16a

Merge branch 'upstream-main' into add-llm-citation-parsing

c353879

InAnYan added 3 commits October 3, 2024 12:17

Refactor plain citation parsing

991e8b8

Merge remote-tracking branch 'origin/add-llm-citation-parsing' into a…

aebd8ac

…dd-llm-citation-parsing # Conflicts: # src/test/java/org/jabref/logic/importer/WebFetchersTest.java

Fix checkers

a981864

koppor mentioned this pull request Oct 4, 2024

Enable search for diff tsantalis/RefactoringMiner#787

Closed

koppor requested changes Oct 4, 2024

View reviewed changes

Merge branch 'main' into add-llm-citation-parsing

b575c96

InAnYan added 2 commits October 6, 2024 14:04

Fix from code review

e55f8d4

Add comments

76c7b8a

InAnYan requested a review from koppor October 6, 2024 11:07

koppor requested changes Oct 6, 2024

View reviewed changes

Comment thread src/main/java/org/jabref/gui/preferences/PreferencesDialogViewModel.java Outdated

Update src/main/java/org/jabref/gui/preferences/PreferencesDialogView…

e1e35f4

…Model.java Co-authored-by: Oliver Kopp <kopp.dev@gmail.com>

koppor enabled auto-merge October 6, 2024 11:37

koppor approved these changes Oct 6, 2024

View reviewed changes

koppor added this pull request to the merge queue Oct 6, 2024

Merged via the queue into JabRef:main with commit daf6736 Oct 6, 2024

koppor deleted the add-llm-citation-parsing branch October 6, 2024 11:51

koppor mentioned this pull request Oct 30, 2024

"New entry from plain text" gives strange results #11805

Closed

2 tasks

ThiloteE added the component: ai label Oct 30, 2024

tsantalis mentioned this pull request Apr 20, 2025

Improve matching of Method references pouryafard75/RM-ASTDiff#111

Closed

Uh oh!

Conversation

InAnYan commented Sep 25, 2024 • edited by ThiloteE Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Mandatory checks

Uh oh!

koppor commented Sep 25, 2024 • edited by InAnYan Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Siedlerchr commented Sep 25, 2024

Uh oh!

koppor commented Sep 26, 2024

Uh oh!

ThiloteE commented Sep 26, 2024

Uh oh!

ThiloteE commented Sep 26, 2024

Uh oh!

koppor commented Sep 27, 2024

Uh oh!

ThiloteE commented Sep 27, 2024

Uh oh!

ThiloteE commented Sep 27, 2024

Uh oh!

koppor commented Sep 29, 2024

Uh oh!

koppor commented Sep 29, 2024

Uh oh!

InAnYan commented Oct 3, 2024

Uh oh!

InAnYan commented Oct 3, 2024

Uh oh!

InAnYan commented Oct 3, 2024

Uh oh!

InAnYan commented Oct 3, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

InAnYan commented Oct 3, 2024

Uh oh!

Siedlerchr commented Oct 3, 2024

Uh oh!

InAnYan commented Oct 3, 2024

Uh oh!

koppor left a comment • edited by InAnYan Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

InAnYan commented Oct 6, 2024

Uh oh!

InAnYan commented Oct 6, 2024

Uh oh!

koppor commented Oct 6, 2024

Uh oh!

koppor commented Oct 6, 2024

Uh oh!

koppor left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

InAnYan commented Sep 25, 2024 •

edited by ThiloteE

Loading

koppor commented Sep 25, 2024 •

edited by InAnYan

Loading

InAnYan commented Oct 3, 2024 •

edited

Loading

koppor left a comment •

edited by InAnYan

Loading