And we'll get this webinar rolling.
Welcome everybody here.
My name is Cody Winchester.
I am the director of technology at IRE.
Thank you for coming to hear about our brand new resource center.
I'm going to be handing this over to Ben Welch here in a minute.
He and, or the colleague, Derek Willis have been working on,
if you have not seen it, I'm going to drop a link in chat, archive.ire.org,
which is our new repository for contest entries, tip sheets,
all sorts of resources.
We are working on supplementing that as we go.
And Ben, you just recently described this as our 1.0 version.
We're looking for feedback.
So please kick the tires and let us know what you think.
And I know this recording will be made available afterward
for anybody who has to leave early or wasn't able to make it.
And otherwise, I'm going to turn it over to Derek,
chair of the board's website improvement task force,
who's done quite a bit of work to make these resources more accessible.
Thanks.
Thanks, Cody.
And thanks, everybody, for being here.
Really, what I want to, the way I want to begin is by talking about,
like, we, I think, all know or like have an idea of like how valuable
the resources that come out of our IRE's trainings and conferences
and contests are.
And like, if you've been a member or been around the organization
or been to one of the conferences, you know that like you go,
sometimes a lot of these sessions you go to for the tip sheets.
The tip sheets are great or the contest entries are great.
You're like, wow, I really want to know how they did that.
And so the effort that our task force and in particular Ben worked on,
has been working on for the last year,
is about trying to harvest that knowledge and that context
and make it accessible to the IRE membership.
And so to that end, like, thanks to the efforts of Cody and other folks
in terms of getting us, you know,
digital representations of conference tip sheets, of contest entries,
we're able to now have a archive site for those and that allows user
and IRE members to search for interesting things, for useful things,
for like, oh, hey, I remember this session from this conference
and I want to find that tip sheet or I want to see, you know,
what's been done on this topic and all the, you know,
all the contest entries that we have and, you know, other kinds of materials
from our conferences, IRE's conferences and trainings.
And so that's kind of what we've built with the archive site.
And so we want to kind of show that off and talk a little bit about
what that entails.
Ben, can you do me a favor and just drop in like the presentation link,
because I think I've misplaced it somehow.
I sure can.
And I am clear of the thing I had to watch out for.
Oh, well, then if you'd like to, you give a much better spiel on this
than I do, but it's up to you. I'm happy to.
Let's, I'll pull it up and then let's do edit like Kenny and Dolly.
That sounds great.
All right. Can you guys see my screen?
Yes. Excellent.
So, yeah, you heard it from Derek. There's a brand new website.
We launched it at the night car conference in Indianapolis.
You can go to it now if you'd like at archive.iree.org.
It's our new resource center.
And I'm going to tell you how it works.
You can see here on the homepage, which is what you'll see after you log in.
When you first go, you should be prompted to sign in because this is members only.
After you log in, you get a big search box.
And if you type in your own name or the subject of whatever you're working on
at your job right now or whatever you're interested in,
you should get back a list of related IRE resources.
Each one, there's about 33,000 that are currently in the database.
And each one of them has its own sort of detail page or URL
where we present everything we have for that resource, which, you know,
has its name, who came up with it, what type of thing it is.
And then in most cases, we have the associated thing itself,
be it a PDF of the tip sheet that was scanned in or emailed over
or an audio recording of a panel or other event.
Those are the two most common things.
And so 33,000 are in there.
What you'll find if you go in there, if you fish through it,
is that the majority of them are IRE contest entries.
So, you know, through the years when people have applied to win an IRE award,
they've had to fill out a form that explains, you know,
why they should win and talks about the method behind their work.
And that actually accounts for the majority of the 33,000
are those contest entries, which stretch back decades.
In preparing this site, I heard from some of our members
who are passionate users of the contest entries.
I can share with you a couple use cases I heard about
from real world investigative people.
One is an editor here at Reuters who told me that he uses the database
to look for and sniff around topics that he's interested in exploring
and maybe investigating.
And what he'll do is look for maybe classic examples
that he hadn't heard of from decades past.
Maybe they're localized in different places
and ask himself, could I nationalize the story
or is there some lesson I can learn from what's been done in the past?
It also allows him to see whether other people have already covered the waterfront
or whether there might be new stuff to look up.
His view is that newspaper archives are so decayed
and the internet is so biased towards the present
that he often can't find the same information when he searches in Google
and that looking at the contest entries gives him a more direct look
at what investigative reporters and editors have been doing.
So that's one case.
Another case I heard was from people applying for awards.
So you could look at the applications from the previous year
and say, hey, who won and why did they apply
and what things did they emphasize or not?
And that can help you write a better package yourself.
That was another tip I heard.
And then a third use case, which I did not anticipate
but makes a lot of sense,
is I think a number of journalism educators use the contest entries
to help their students learn about the methods and practices
of kind of like the great stories in IRE history.
So if you're having your students study the Boston Globe
child sexual abuse story
or pick your greatest hit of investigative journalism,
if you find the contest entries for those stories,
you can often get a lot of insider detail about how the data analysis
was done or the story was put together, the kind of behind-the-scenes stuff
that could be really helpful for students to learn.
So that was the third interesting use case I heard about contest entries.
The number two most common thing in the database are, of course, the tip sheets,
which I suspect are the most popular.
I don't have data to back that up, but that's going to be all the materials
presentations through the years with the PDFs attached.
We have 6,000 of those assembled there in this data set.
There are a few duplicates and kind of reruns from year to year,
but it's pretty fresh, pretty fresh and pretty unique.
And then the third large group are the audio recordings.
These are all the panel discussions and events that happen at IRE conferences.
A large share of those are put down to tape,
and you can listen to MP3s of Bob Woodward and Leonard Downey
talking about this or that,
or Sarah Cohen schooling you up on the database frame of mind way back when.
You also are able to listen to the audio recordings of more recent events as well
and find stuff related to topics you're interested in.
And for me, as someone who was really exploring this archive in depth
for the first time making the site,
I had a lot of fun going back and listening to some of the older stuff
and tuning into the golden oldies.
Every asset in here has a common set of metadata
or columns in the spreadsheet that go with each record,
and some are filled in better than others.
This is not a perfect data set.
This release we're doing is really a first effort
just to get this stuff up and out,
and there's a lot of room for improvement.
And this is one of those areas where we think we can improve.
You can see why everything has a category and a title and description.
Our author's field isn't very well structured or formatted.
It's a little loosey goosey.
And we've got some pretty big gaps as far as the year or date
when the records were created,
which means that when you do a search and try to sort
for what the most recent document is,
which is a really common use case we heard from people,
we don't have a perfect solution yet.
And that's something we want to get better at
bringing up these numbers that you see here on this chart
is one of our goals,
and we want to gradually add more and more extra stuff on as well.
I will jump in here to say that this is much more
of a problem of a past resource problem.
The efforts of the staff have really helped us have more data,
more metadata information about more recent resources,
tip sheets, conference entries, things like that.
But there are quite a few resources from older conferences,
events where that information isn't present.
And oftentimes that affiliation, which is one of the lowest ones,
essentially that's being extracted from people's names
and information on, say, a tip sheet.
And in those cases, if there wasn't really a standard,
people might just put their names and emails on a tip sheet.
And so there's literal cases where the information isn't there
and we have to either sort of derive it
or maybe we won't get it,
or we have to wait until someone might claim it.
Yep.
And this database that you can search with uses
what's called semantic search.
And so this is kind of an AI-flavored version of search
that goes a little beyond just doing kind of strict keyword matching
of what you're putting in.
Kind of under the hood, inside the database,
we've had a machine learning system kind of break up each asset
into kind of its component parts
using the kind of fancy vectorization and AI stuff
we hear so much about these days.
And that allows it to take your query, what you search in there,
and find something closer to what you really mean.
And it also allows us to sort of rate documents
based on their similarity to each other
and other interesting things.
All of the code that does this is open source on GitHub
on the IRE apps profile page, if anyone's interested.
And you can see the Python code we do.
I'd be happy to get nerdy in the question and answer portion
and go into the details of this,
but I don't want to bore you too much.
But we can brag and say we do have an AI-enabled search,
which is pretty cool and was a big learning experience
for me putting this together.
After you do your search, you see a result page kind of like this,
which takes the result of that AI funnel
and says, here's what we think are the best matches for you,
which you then can pick and see from the list how you like.
You're able to tune that search by, say,
limiting it to different categories.
Say you only want to search tip sheets or audio.
You're able to just check and uncheck those boxes and do that.
You're allowed to sort newest and oldest in addition to relevance,
but it is imperfect due to those gaps in the data currently.
And if you really, really hate AI,
you can just turn it off there with that little switch
and do a standard keyword search as well.
At the bottom of each of pages, each assets page,
you can see some recommended similar documents.
This is based on that same AI sugar where we're able to say,
you know, this one is most similar to that one and on and on.
I did have an AI bot make these similar scores for my slide,
and I can see that it has titled one,
a guide to ass kicking in the 90s.
I just noticed that.
I don't think that's a real tip sheet.
Maybe it is a real tip sheet. I don't know.
We'll have to look that one up.
As we said at the start, this is member walled.
So just registering with the IRE website is not enough to access this database.
You have to both be registered and have an active membership,
aka have paid your dues in the last year.
If you don't do that, you can't get it. Sorry.
And this is the kind of software tools we're using.
We're using kind of a vectorized database,
which is a newfangled thing to kind of build.
And this site, I began, I'd be happy to talk about the tech later
if people want to get into it.
But, you know, just to wrap up,
I just want to say this is a work in progress.
It's a very simple search that has a couple interesting features,
but there's so much more we probably can and should be doing with this.
We want to hear your feedback on what's broken or not like it should be.
We want to hear your feature requests on what it should be doing, that it's not.
And if you're really passionate about this,
we would love to welcome you into our volunteer creation process.
We have a sort of list of priorities going forward.
Our top priority is to expand this system to make it possible for the IRE
staff to do rolling incremental updates.
Currently, we're just able to do like big database dumps every so often in
that requires like me or Derek to push a button or Cody.
And we'd rather have like an administrative panel
where the staff can be actively improving it.
That's our top priority.
We're going to do that in the coming months.
But then we also have some gaps in the data,
stuff that was lost in the migration, some of these flawed fields.
We want to try to start cleaning that up.
And then finally,
we have some really interesting stuff associated with the assets.
We haven't integrated yet like the OCRing of all the documents,
the transcribing of all the audio files.
We do have all those associated transcripts,
but we have not yet integrated those into the search or the interface.
You can't see them on the site.
Your search doesn't benefit from them.
And we'd really like to get around to that before too long too.
So that's kind of our roadmap.
The whole thing was done on a volunteer basis by really the three of us
and our colleagues on the volunteer task force.
So be nice to us because this was a nice weekend thing.
But we definitely also want your brutal honesty about how it can
and should be better as well.
And that's really my spiel.
And I don't know if Cody has an agenda beyond this,
but I'd love to hear from folks of what they think about it,
how they use the IRE archives and how we can make them better.
No, let's go ahead and open it up if people want to chat.
Okay, I can get us started.
Hi, everybody.
Thanks so much for sharing.
Well, for building the tool or for sharing it with everyone.
One piece of feedback I gave in the Slack that I wanted to follow up on
was categorization or the possibility of categorization.
I had mentioned that for me as someone who is new to journalism,
I'm not aware of what I don't know.
And so even coming up with what to search for is hard
when I'm presented with just a search bar on the homepage.
So definitely something like that would be really helpful
as a future request for me.
You'd be interested in kind of browsing topics you're interested in.
Am I hearing that right?
Exactly.
Yeah, we've heard that you're not the only person who said that.
You know, we had several other people at NyCar raise the same question.
They're kind of like, I want to nose around.
I want to walk through the stacks a little bit, you know.
And I could imagine, you know, we don't currently have kind of tags
or categories beyond the sort of like, this is a tip sheet.
This is a contest entry level stuff.
And so getting that, like applying those groupings
is like something maybe AI can help us with or I don't know what.
Like, so that's a challenge.
And we were actually talking about that earlier today.
Like, how could we do that in a disciplined way?
A second idea we've heard, I wonder what you would think of this,
is maybe having the IRA staff or muck rock or members of the community
curate kind of like maybe greatest hits kind of lists,
like a reporter who really knows a topic or someone on the staff
collected the 10 best tip sheets on topic X
and it was a little more of a DJ playlist.
Does that appealing or how do you react to those ideas?
Yeah, I mean, I would definitely love something like that.
I think if those curated lists could still also link to the non,
the top, you know, top 20 or 10 to 20, 20 to 30 of whatever else is left.
So I could then follow up with other things.
I think that would be really helpful.
But as a place to start, for sure, I would love to see something like that.
Yeah.
And I'm assuming it's also tricky because language changes over time too.
So the way something is described in the 90s maybe is described with using
different words now.
So I'm sure it's a challenge.
That's an interesting question.
There's a PhD thesis for you, Derek.
I mean, there's a lot of, I'll pass, but there's a lot of like,
there are definitely sort of topics or specific phrases or techniques
that kind of die out for not extinction level stuff,
but in terms of like what IRE members talk about.
Yeah.
It's like as if that were the case that essentially like we used to talk
about these things and now we don't talk about that specific thing anymore.
We've replaced that with something else or we've, we've moved on.
So there's definitely a, a life cycle to some of these, some of these things.
I, I, you know, not like mostly not for things like if you're browsing by,
like I want to do the story on this kind of institution or things like that,
or I'm interested in what's been done in this area.
Like probably that's fine, but like there's definitely some specific
technique stuff that is either, that is basically now mostly dead in terms
of like we have, like we have a lot of people talking about it, but like,
like there are a lot of tip sheets in there on how to search the internet.
Right.
And right.
And now maybe somebody would use the word OSINT if they're talking about
talking about like, yeah.
And so like, and so like, but I mean, I do think that like Ben,
like the semantic aspect of this search engine does handle some of those
use cases where like, if I'm describing what we now call OSINT probably
should find that stuff anyway.
Right.
Or a car versus data journalism, that kind of thing.
I think it's like, like that's, that's where like, that's the whole,
that's the whole ball game for semantic search.
That's why you use it is to be like, I.
Yeah.
Like it could appear in all these different ways.
So.
Yeah.
Cool.
Thank you.
Um, once I could follow up one small bug that I have noticed so far is if I
click on, I haven't pulled up, if I click on an item and I see that there's
a category tip sheet or audio, if I click on it,
it just takes me back to the homepage search bar rather than I was,
I was like, I was expecting, I guess I would say like a list of all the audio
ones or all the tip sheets kind of in one spot.
That's what it should be doing.
I mean, yeah, like, like, oh no, no, she's absolutely right.
Like I just tried it with contest entry and it just goes right back to the
homepage.
This is like one of my Svelte kit bugs.
Yeah.
Where like, you know, you're trying to do that.
I don't want to drag you into my JavaScript life, but like,
this is something that like where it'll work on your local development.
And then not when you build it out on versatile Derek and like,
there's like little things like that, that I, that it's good people point out.
So, wow.
Should we just like assign AI to do this right now?
Get on it, Claude.
I mean, probably.
I'm doing it right now.
Other, other thoughts.
Yeah.
We went here.
We want your bug reports.
Thank you so much for that.
That's really helpful.
Thank you.
Thank you.
Thank you.
Thank you.
Thank you.
I'm curious to know if folks have thoughts about how we should,
or whether or how we should handle.
Not whether, or essentially how we should handle.
We have like, we've got some amount of non-English language.
Resources, right?
Yeah.
We, we, we do have some number.
They're not currently categorized that way.
And a lot of the.
A lot of the backlog that we have to load into the database from the last
couple of years, we have quite a bit more mostly Spanish language content.
But being able to categorize by language is another.
I mean, when we're, when we're talking about, you know,
What does the process look like when we're loading new resources into the
database, being able to categorize by language is one of,
is one of our, on our wish list.
Okay.
Other thoughts or.
Or future suggestions or.
We'll make this more useful.
Beyond brow, beyond browsing, which we definitely now hear.
Silvia, I just linked you to the issue where AI has been assigned to pick
it up. If you're interested in.
Co-pilot as a spectator sport.
Let me know if I got the description of the bug wrong. I think I got it.
But yeah.
Hi.
Hello. Hello. How are you?
I'm great. Thank you. And thank you for organizing this.
I found really useful.
I don't know if it is already possible or covered,
but I was wondering if.
There might be like a connection in database.
If a story won an award,
one of like a Pulitzer award or,
or another award. And so you can have that.
Like the, the most important,
the category of the awarded stories.
Hmm.
That's a good, like,
that's a good idea for one of these curated lists, right? Or you, you know,
you would say here, here's like,
here's all of last year's winners, right?
Maybe you're something like that or.
And then.
Sorry. And then it might be interesting.
If the authors or someone else will write.
A behind the scene.
A story of the story.
About the tools used.
The method.
The methodology and the tools.
These will be the real value.
I think I don't know if the other.
Agree.
So we do have like, well, I mean,
first of all, we, we kind of used to have for,
for like technical sort of methodological stuff.
We, we.
I already sometimes will.
You know, obviously there's,
there's some panels and sessions about that that go into some detail
about that.
This is also what we used to use uplink for in a lot of cases.
Or in some cases, at least a description of it.
Uplink was like this older newsletter that we had.
And.
And I think we're going to be able to,
at least one of my goals is to be able to integrate the,
the text of uplinks into there.
As well.
But I do think that like.
There's a.
Like.
Like the, the, the, I think the challenge,
one of the challenges for that is that there are like the tip sheets
are usually like a summer and even the conference presentations
themselves are usually a summary of that rather than like a play by
play.
So.
Oh, we're bringing back.
I read your.
Oh, all right.
Nice uplink returns. Great.
Great.
But I do think like one of the issues was sort of like, I guess it
depends.
How detailed do you want that methodological stuff?
To be.
Because a lot of times we don't get all,
we don't really have all of that as part of our resources.
Okay. Thank you.
It might be interesting also to involve the voters of the,
of the stories.
So we would like,
we would like people to be able to like members to be able to claim
resources that they've uploaded.
You know, it's sort of been like, you know,
if you go like to a member's page,
the member directory and be like, Hey,
I've got here are the resources from this member.
And like, I, you know,
eventually allow them to like add thing or at least, you know,
augment that.
This information that would be,
that would be a great thing to do because I suspect there were people
who have things that are like, okay.
Yeah. Most of this advice is still good, but this thing is not.
Or like I've updated this here or like, you know,
you should contact me about this because things have changed.
Thank you so much.
Yep. Do you use the, um, the resources yourself much now?
I will.
Okay.
I'm curious what people are doing in real life.
You know,
cause you make something like this and it's hard to know like,
what are people really doing out there? You know,
I wasn't aware of these resources, but now,
thanks to this webinar, I'm, I'm aware.
So thank you for organizing it.
Of course.
Because we receive tons of emails.
So, and sometimes I don't even open and just read the title.
So, because there are so many emails and we need to prioritize.
But so these, there was in the title,
in the headline of the email, there was written the new resource center.
I said, okay, interesting.
Little editing advice there. I love it.
Yeah. Other thoughts or questions.
Yeah. Well, if nobody has any other questions or sorry,
Sylvia just saw you unmute. I didn't mean to interrupt you there.
Oh, I was out of curiosity. I was wondering,
so every time you refresh the homepage,
there's a different set of suggestions under the search bar.
And I was curious how those are generated.
It is a list that I made up and they are drawn randomly.
And do you have one you'd like to see added?
We can make a pull request right now.
Yeah.
We just add to it.
I don't have anything particular, I guess, but I do often rely on like,
as far as browsing, sometimes I will look there to see like, oh,
what's this other topic that I know nothing about? And I'll click on it.
Yeah. Yeah.
It was me trying to be cutesy. So if you click this link, Sylvia,
I don't know if you're interested in JavaScript,
but you can see the full list right there.
So I wanted them not to be too long so that it wouldn't wrap the two
lines. Yeah.
I do think that we also have the opportunity to possibly like, you know,
maybe we do something where it's like, um,
all like in the month or two or, you know,
after a conference where it's like one of the canned searches is tip
sheets from night car, this year's night car tip sheets from this year's
IRE conference, whether we could do something like that.
Yeah. I think that would be really useful too. Mm hmm.
Cause so many people ask for the chip sheets like right after the
conference you're looking for. I'm like right after. Yeah.
So the constant annoyance of our staff, I'm sure.
We're like, yeah, we're getting there. We're getting to them.
We got to upload all these things.
Yeah. Steep sheet and the presentations,
or if you can share all the presentation, that would be great.
Yeah. Those will be included in many cases. Right. Cody,
if people had like a PowerPoint deck or whatever.
Oh yeah. And in every case we, we have, uh, you know,
we, we talked a little bit about getting some of the, um,
resources from recent conferences.
So if you go to that list that I posted in a chat,
all of those are eventually going to get loaded into, um,
the resource database plus a bunch more, um,
you know, kind of odds and ends. And then, you know,
figuring out like how, like webinars like this, you know,
we record it. Uh, we usually upload this like to our Vimeo, uh,
account.
Our, our previous, uh, system did not have a concept of like,
you know, an embed code for a video or even just like a link to
something as opposed to like a document.
So this will be a lot more flexible and we'll be able to, um, um,
yeah, anything that the speaker send us plus the audio that we record,
we will upload into this for sure.
Thank you.
Any other thoughts, questions, ideas, bug reports?
So be asked what help will we're looking for? Well, bug reports is a big one.
Feature request is another one. You know,
um, for people who are interested in programming or the kind of library
science of doing this kind of thing.
We'd love to welcome people in. The whole thing is open source software.
Um, and that's not just, uh, to show off.
It's really a desperate plea for help.
And, uh, so, you know, if you also have,
if anybody has time or interest in improving the software,
we'd love to welcome you in.
Um,
if anybody has time or interest in improving the software,
we'd love to welcome you into that.
We're currently have a closed source repo where we're trying to work on
that administrative panel. I mentioned, so the staff can update it.
That isn't open,
but if people are interested in participating in that software project,
I would be happy to welcome them in as collaborators.
Um,
and I also, yeah, I think also another way is like,
because we've been talking about,
and then heard from people about like the aspect of like browsing,
like we have our ideas about that. Obviously, you know, like those,
right? Like not randomly, but the ones that Ben already has put in there,
uh, suggestions, but like if people have ideas for like, okay, well,
if I were going to design like a browser, you know,
like a topic system here, uh, you know, like, what would that encompass?
Like, what should that, what would that look like?
What should that look like? Um, like,
I think the more suggestions and input we have on that,
the better any system that we might come up with is going to be.
And I just, uh, toss in the chat, if you just email,
and this is on the, the, uh, page as well.
If you see an issue, email help at irie.org.
If you have a feature request, that's a great place for it too.
All that eventually gets filtered over into a, you know, kind of a bug,
bug track, bug and feature tracking websites, uh, that we use.
So definitely, uh, please kick the tires and let us know what you think.
Yep.
Also that the AI has already sunned me and, and, uh, made its suggestion
for how to improve the code.
So you can see the link.
I've got the, the pull request is already in, I,
what it's describing as the bug sounds right,
but I kind of want to pull that down to my computer and look at it first
before we send it out to the world.
But, um, I think that's probably a pretty good guess just looking at it.
I probably just made the link Ron.
Or I'd made it one way and then changed how it works and forgot to update it.
The kind of dumb stuff you do when you make websites.
Anybody else?
Final thoughts, questions, discussion.
All right.
Well, um, thank you so much to Derek and Ben for being here today and also
just for all the work y'all have put into this.
It's fantastic.
It's such a huge improvement over what we have.
And thank you all for coming to give your feedback.
Like I said, um, help at ira.org, um,
is going to be your email address if you have any feedback.
Um, otherwise we will see around and we will post a recording of this,
uh, on our website soon.
Thank you everybody.
Thanks for coming.
Thanks for your thoughts.
Yeah.
Yeah.
Thank you.