ERDWMeet20220707

Summary from: Eric Zinda

In attendance: Olga Zamaraeva, Dan Flickinger, Luis Morgado da Costa, Alexandre Rademaker, Francis Bond, Eric Zinda

Great first meeting! Summarized below. All actions taken are preceded with "Action:" below. Search for that to find out what changed.

Next Steps

Names in italic are the owners for that item.

(for 7/14) Deliver 1st Prototype of:

• (eric – alexandre - olga) Conceptual Doc
• (eric – alexandre - olga) Inventory Doc
• (alexandre) Review Perplexity Case Study Doc and give feedback on it as a Case Study Prototype
• (francis > 7/14): Deliver another case study prototype for: machine translation

(for 7/14) (Dan/Luis/Francis/Olga) Deliver a list of what we already have for:

• Conceptual Docs
• Inventory Docs
• Case Study Docs
• Tools Docs

(Luis): Send pointers to docs for full forest tree bank (and what to do with it)

(Eric): Set up a meeting for next Thurs: 11am. Agenda:

• Go through prototypes from above
• Discuss what approach to take at the conference
• Survey of existing stuff

(Action: Created ERDWActiveWork so we have one place that always lists current deliverables)

Meeting Summary

We went through the attached slide deck and had a high level discussion about what we wanted to accomplish.

(Action: Created ERDWMeet20220707 to summarize the meeting)

(Action: Created ERDWCharter to summarize what we are doing)

(Action: Created ERDWIssues to collect the working group open issues)

Audience should include Grad students (and people like them)

In the past we have people from graduate programs using the ERG output and modifying the grammar. They are different from comp sci folks in that:

• They have had a mentor in delphin
• They have a linguistics background at some level
• They also have a comp sci background

The group noted that these folks would appreciate/want/need things like:

• Journal references
• Background on the "why" of things that are there
• The linguistic technical terms, etc.
• Documentation on modifying the grammar itself (as opposed to using the output)

much more than comp sci folks. It was proposed to consider them in the intended audience.

On further reflection I'm going to split this into two things (actions described at top):

1. They want practical grammar editing documentation. (Action: Added "grammar editing documentation" to the content section of the Charter)
2. They want more of the linguistics-focused content in all the reference docs we write (Action: added to open issues list)

Audience should include many different kinds of researchers (not just linguists)

The group recognized that part of the audience we should target are non-linguistic researchers too (e.g. logisticians). Looking through the Charter, I feel like we have already captured these folks in the audience by saying:

• Anyone who wants to build a system using the ERG
• Must have computer science experience

But it was good to clarify this group.

Get a good a survey of the documentation work that has already been done

The group noted that we need to better understand what has already been done like:

• Francis work with web interface to browse grammar
• Annotations in the source
• The Wiki
• Existing papers
• etc.

(Action: added to open issues list)

(Action: Created ERDWDocSurvey to collect references to all the docs that are already there)

Outside of MRS and Syntax, we need to document "extension points"

We had a discussion about how there are many parts of the architecture that could be ripe for enhancement and exploration by users, if they only knew about them and had documentation for how to "hook in": parse ranking, lexical filtering for improving efficiency, ? context free approximation?, were brought up as examples.

(Action: Added "Extension Points" to the list of types of Content in the Charter.)

Other discussion points

Changes to grammar rule names: irritating but inevitable. Give some thought to the fact that underlying things will change. Find a way to alert people that changes have happened. Release Notes? (Action: added to open issues list)

Dan noted that you can make radical changes to the syntactic structure without changing the public interface (MRS). We should treat semantics as more permanent even if the inside changes.

(SMI) - shows the abstract semantic predicates and lists all of the potential predicates in the MRS. Users should expect these things to come out with a build. (Note that it is already published with each release). (Action: added to open issues list)

Internals of the syntax effort: This group needs to understand better the lexical type database, the mechanism that francis and colleges have built. How to tie examples with example sentences. There is more work that can be done to exploit that tool better for grammarians. (Action: added to open issues list)

If things go well, look at applying at googles season of docs to get funding and supervise them. (Action: added to open issues list)

Two issues from the past: - Having the doc strings in the grammar has worked well in principle but in practice we can't share the burden beyond Dan. Is there a branch of the ERG where people can put in docs and it gets reviewed? - How large is the pool of people that can document the semantics of the ERG? (Action: added to open issues list)

Search is a big issue (already in issues list)

How to getting visibility into the current plan for the ERG? (Action: added to open issues list)

Alexandre pointed out that a key part of enabling a wider audience to use the technology may include updating the tools to make them more stable or more familiar to a broader audience. Moving the Wiki to github was brought up as an example.

(Action: Added a section on "Modernizing Tools" to the Charter. Right now it is just about collecting and prioritizing the list)

(Action: Created ERDWToolsFixes to collect tools enhancements for broader audience)