OEWG 20090429

From GO Wiki
Jump to navigation Jump to search

OBO-Edit Working Group, Wednesday April 29th, 2009 9.30 a.m. PST

Conference call details:
US: 1-888-727-6732
Outside US: 1-719-867-3417
pin: 601425

Agenda/Chair: Melissa
Minutes: Amina
Attendees: Midori, Jen, Melissa, Tanya, David Hill, David Sutherland, KarenC, Nomi, Harold, Thomas, Amina

Action items carried over

  • Action items: Amina
    • Finish work on assert-implied-links component, and email OEwg list
  • Action item: Midori
    • Inform a few others who might be interested in the demo
  • Action item: Jen
    • Visit Colin's group, watch users, try to solicit bug reports, etc.
  • Action item: Working Group
    • Prioritize a bug and feature tracker
    • Pick one bug you would like to see fixed by the next meeting

Discussion items

Amina presents: Assert Implied Links Component Demo

  • feedback1: Undo for links asserted action
  • DavidH: To undo - delete link and run reasoner again
  • DavidS: what do you do with these hard links? I'm less trusting of the remove redundant links calculations
  • Midori: existing remove redundant links behaviour: list - click to check what its found
  • feedback2: Reverse select term- link from OTE to AssertTable

DS: multiple problems with this
previous releases of the reasoner - over-flagged redundancy
the only one I trust is RBR
did a lot of checks - and works alright
will be interesting to see how the two different approaches work out:
GO approach: manually assert links and manually trim redandant links.
fly anatomy approach - trust the reaonser and assert all implied links / strip redundancy automaically only as part of a release.

Amina: how will incremental reasoning fit in to the assert links? DPH: yes; assert 1rst, turn reasoner off, then on again to look. But need to wait for Chris's script to run the make XP's. Incremental appears too slow. David O has to always have the reasoner running. GO doesn't; they edit first with off, then turn on to check.

DH: run reasoner again
unless we start creating cross-products ourself - we need to run chris's script
we'll create a xp and then run the reasoner again

DS: Im afraid I'll hit the limit so I turn off the reasoner
MH: for years we've had that as a goal for GO to edit with the resoner on
All: its just really slow
MH: you can have 20 cups of coffee waiting for it
DS: switching the incremental reasoning off should solve this issue

Does remove redundant links work with rule-based vs link-part reasoner. DavidS says it didn't before with link-based. Midori: it does give you a list first so you can check the ones you want removed. David O trusts the rule-based reasoner to not make errors.

Thomas: explaining the ontology generation plugin - presented at the biocurator conference (documentation available in OBO-Edit User Guide) Jen: can we have 3 separate tabs for the 3 steps

Verification Check issues?

background is explained in an email from Karen:

Date: Tue, 28 Apr 2009 14:46:30 -0700 (PDT)
From: Karen Christie <kchris@genome.stanford.edu>
To: GO Ontology Editors <ontology-editors@genome.stanford.edu>
Cc: OBO-Edit Working Group
Subject: [OBO-Edit Working Group] non-critical warnings on GO file

Hi GO ontology editors,

I did an edit today and noticed that there were 26 non-critical warnings. 
I went through all of them to see what they were. There are a couple types 
of warnings where we should probably change what the verification check 
looks for (which is why I cc'd the OEWG on this), but there were a bunch 
of user errors which people should have fixed before they committed.

In the process of releasing OE2, Jen did a lot of work to clean up the 
hundreds of these that we used to have, so that people could actually use 
the verification checks to catch problems they introduced. But if we start 
collecting a whole bunch of these again, then everyone will ignore the 
verification checks again and we'll be back to where we were, and 
eventually someone will have to go through and clean them up again.

I think it would be best if we can keep GO "clean" of these types of 
problems so that the verification checks are useful to each person as they 
save, so they can use it to fix their own problems BEFORE they commit 

Below is what I found in going through the warnings. Maybe we can talk 
about appropriate procedure to avoid accumulating these warnings, and 
perhaps the OEWG can talk about whether two of the checks are picking up 
things they shouldn't be.



1. User Errors: Almost half were simple typos, e.g. "anaphasep" instead of 
"anaphase.", internal newlines within definitions, or missing final 
periods from definitions, the latter often occurring in defs from EC or 

It seems that people should NOT be committing the ontology with these 
types of errors, they should fix them before they commit so that we con't 
accumulate scads of them.

There was also one url in a definition. By comparison with the other urls 
that the verificatino check flagged, it seems that perhaps this should be 
in the comment, not the definition?

2. Verification Check issues:

Then, there were two other types of warnings, where it looks like maybe 
the checks are picking up things that should be allowed.

Repeated word - There were four "repeated words" reported where it ignored 
the fact that there was punctuation in between the two instances of the 
repeated word. While two of these might be less than gramatically ideal to 
use the same word twice in close succession, none of these are illegal, 
and two of them there is probably no other way to phrase it. Perhaps it 
should not report repeated words when there is punctuation in between.

Issue with sentence boundaries - Most of the rest of the warnings were 
about periods with no whitespace after them, resulting in two warnings:
- sentences that do not start with a capital
- sentences that are not separated by whitespace.

However, none of the flagged issues were supposed to be sentences. Most 
were urls in comment fields. A couple others were names or formulas that 
contained periods where there was no whitespace after the period. Perhaps 
we should not look for periods followed by a non-whitespace character.

Conclusion: Relaxing the rule for verification marks will be easier than maintaining files listing allowed exceptions

Short term fix:

  • Action Item - Jen: fix period words and repeated words files and commit to svn

Long term fix:

  • Action Item - Karen: submit feature request for changes to verification check rules
  • Action Item - Amina: edit verification check rules, pending decision on relative priority with respect to time required and other priorities

Harold asked about whether OE2 was producing warnings during text edits. Midori commented that in some previous versions, there had been a bug where the verification checks were not actually being done, even if you had selected the options to have them done on text edits.

other topics

  • ID's in xps tab in Text Editor
  • ID where there isn't a name available AI- Amina
  • Tough ontology to test: chebi reasons well

Upcoming Demos

  • Incremental reasoning
  • Parent Editor


  • parent editor
  • Potential Parent Editor Improvements: show which links are asserted and which ones are implied
  • doesn't need anything else
  • Discussion b/w Harold and Alex

(Idea: ability to save the non-fatal errors list to a file. Also, discussing this with Alex: why do we care about extra white space since it's non-fatal. What will it harm?)

Action Items

  • Amina
    • feedback2 for Assert Implied Links Component: Reverse select term- link from OTE to AssertTable - useful when looking at a long list of asserted links
    • 2.1 release
    • bug fixes
    • Graph Editor Crashes
    • Config Manager memory settings not in sync
    • Look at Graph Viewer again and try to figure out what's going on
    • Look into extended character support

Jen: update period words and repeated words files
Karen: file feature request Midori: check if the "on text commit" verification manager check is fixed
Jen: visiting Colin's group


Back to OEWG Meeting agenda and minutes List