GraphStuff.FM: The Neo4j Graph Database Developer Podcast

Nodes 2023 Preview with Melly Beechwood

Episode Summary

Today, we have a remarkable guest who's about to take us on a captivating journey into the world of data and wildlife conservation. Joining us is the brilliant Melly Beechwood, a seasoned Machine Learning Engineer with a profound passion for data exploration and a unique mission in the realm of knowledge graphs. Get ready to be inspired as we delve into her fascinating work and her upcoming talk at NODES 2023.

Episode Notes

Episode Transcription

[00:00:00] Alison Cossette: Welcome back graph enthusiasts to another exciting episode of GraphStuff.FM. I will be hosting today. My name is Alison Cossette and I'm joined by my cohosts, Jennifer Reif and ABK, how are you guys doing today?

[00:00:14] Jennifer Reif: Good. How are you?

[00:00:16] Andreas Kollegger: I'm doing well. I'm losing my hand in the virtual background, but otherwise I'm good.

[00:00:21] Alison Cossette: We are joined by our very special guest who we're very excited to have on board, who is going to take us on a captivating journey into the world of data and wildlife conservation. So joining us is the brilliant Melly Beechwood, a seasoned machine learning engineer with a profound passion for data exploration and a unique mission in the realm of knowledge graphs.

So get ready to be inspired as we delve into her fascinating work and tell you a little bit about her upcoming talk at NODES 2023. Melly, welcome.

[00:00:56] Melly Beechwood: Thank you so much, Alison. I'm so excited to be here.

[00:01:00] Alison Cossette: Golf claps all around. We're super excited to have you. So, you are joining us today because as we were going through different talk submissions, I will tell you when I got to yours, I stopped and I was like, Ooh, this is really interesting. I definitely want to make sure that I vote this one high.

I wondered if you could tell us just a little bit of a teaser about what you're going to be sharing with us at NODES this year.

[00:01:25] Melly Beechwood: Yes. So I will be giving a talk around how do you build knowledge graphs for wildlife conservation? And the heart of this is my own journey of doing this for the orcas in the Puget Sound, which is in the Pacific Northwest of the United States.

The orca population here has been decreasing for quite a long time, for a variety of reasons, but they're all endangered at the moment. And I really wanted to see if there was something I could do to help them. In that journey, you get to see a lot of how you can pull in multiple data sources to create this knowledge graph and how we can give unique insights that kind of help combine a lot of what other scientists are doing and citizen scientists in the area to really impact and grow their results and see those hidden connections, which I think also makes it a lot easier for potential policy change in the future.

And I'm hoping that this talk will not only show people the importance of looking outside the traditional tech fields, but also for their favorite animal that might be in danger, to have more people go out there and potentially find some insights that could impact some change.

[00:02:41] Alison Cossette: So I have a question for you. Were graphs something that you were working with previously? Like, how did you come to have graph as something that you were diving into?

[00:02:51] Melly Beechwood: Sure. That's a great question. So I had actually been previously working at Tableau and they're really known for their data visualization, but not in the graph form.

My team specialized in refactoring. The, like, legacy code in the area as at the time. I wasn't a machine learning engineer yet. I was just beginning on that journey and I was just a good old software engineer. I was learning about Neo4j, had stumbled on it and had been just been playing with the database.

And was like, I wonder if I could use this to, like, fix my problem with semantic analysis on the code base and see those little connections. So I could see if I made a change there, how, like, is it going to be like a multi month project refactoring or like a two week one? Cause this was driving me crazy.

So I got greenlit by the company to use a certain amount of my time to learn Neo4j and try that out. It actually was successful. It did help a little bit before unfortunately priorities changed when they got bought by Salesforce, but no biggie. It, like, ignited my love for graphs, and right around that time, I started going into my master's degree for machine learning.

And that's actually what the start of this whole orca thing was, is I wanted to look more at that combination of using Neo4j plus machine learning, and what could I do with that? And so, the OrcaGraph actually started out as a tiny side project for a class, and then grew to a semester project, and then is most likely going to be my master's thesis. And it's just really grown from there, but it's been a lot of fun.

[00:04:36] Alison Cossette: Sounds like it. So tell me a little bit about, were there some, were there any sort of pitfalls early on or anything that you learned? Because I know lately we've had a lot of interest from different kinds of social scientists in different ways that they can leverage knowledge graphs.

We're going to talk about some of the technology people are using to access that, but it's been this recurring conversation I've been having with social scientists. So do you have any sort of advice or insight on that road for folks from a more traditional research background that might want to be moving in this direction?

[00:05:13] Melly Beechwood: Probably one of the first ones is most of them are going to be comfortable using at least a basic database and really like stepping aside that fear that yes, this is new, but because of you guys' use of Cypher, it really is not that hard of a step between knowing how SQL databases work and being able to use a graph database.

So kind of like, I think that fear of entry, because it's so big and so new can can be intimidating. And then my next biggest one for all of the biology nerds out there and chemistry, really anyone who works in the natural system. The knowledge graph, like, graphing is not any different than what you already do.

It's exactly the same, you know, whether you're looking at the family genus species, whether you're looking at how molecules connect together. I really think in the, like, in the social sciences or psychology, how people interact. Like, this world is going to be so much easier for you to get into than a traditional database... than using Excel, than using R or any of those kind of tools. I think it's going to fit your brain better, but you have to put aside some of those knowledge, like all of those tools that you've been forced to use, and think about how you already think about the world. Think about those connections.

And you'll be fine.

[00:06:28] Alison Cossette: Gosh, I think that's such an interesting point too, because even just, I mean, I'm a long time educator and even initially teaching folks about relational databases and trying to get them to understand their ERD before they even get started. What I love about graph is how organic it is, literally .

But it really does, to your point, reflect that natural order of relationship of entities, but also just the way our brains work, right? If we think about even just a train of thought, we have a thought about something and then it relates to this and it relates to that. And so I think you bring up a really good point about how, what a natural shift that is to think in graph terms rather than relational terms.

[00:07:16] Melly Beechwood: I think so. And I think the hardest part is when you're so used to the industry forcing you to think in those relational terms of being able to unwind that and come back. And then that joy of being like, Oh, I just put in what I actually think. And I think that's it. It helps a lot.

[00:07:35] Alison Cossette: That's a good one. ABK, I see you nodding. What's on your mind?

[00:07:39] Andreas Kollegger: Every phrase you've been uttering has been like magic. Like, yes, you're going to think the way your brain already thinks. All these things. I love hearing that. We, of course, amongst ourselves, We kind of say this all the time, but like, we don't know, is this just our little echo chamber of the world?

But like having, having you had come to that same realization. And then I think spot on for like your observations on like how fitting it is for the way people are already doing their work, you don't have to kind of twist your brain and figure out, okay, how does this go into a model? How do I actually save this in the database?

Just whatever's in your head, stuff it in the database. You can get it back later in the same way. And it's exactly, and just keep doing your work. Right? It's great.

[00:08:17] Melly Beechwood: Yeah. And it's easy. I think one of the other tips I have. is one of the things that really helped me on making trace the clues, this orca graph is having a label in it with a versioning, if I was going to do something crazy.

Because then it was really easy to search and murder all of them, if it went really wrong.

[00:08:38] Alison Cossette: Oh my gosh, that's genius.

[00:08:40] Melly Beechwood: And then, like for instance, I tried bringing in, I tried using the AWS Entity Extraction, and I also used the Google Entity Extraction. And sometimes that went great, sometimes not so much.

It was very interesting. I'll tease that. There's some funny things there. But that's what I did. I had a little label letting me know, hey, this one was going to be from AWS. This one is from Google. These are my base ones. Please don't like, let's not take those ones out that I did all manually.

[00:09:14] Andreas Kollegger: So in addition to actually labeling entities, you also had extra labels for the source of those entities? How often did they kind of align versus come up with crazy different, like how divergent, I guess, was that where those sets?

[00:09:31] Melly Beechwood: So because of the type of data set it was, when you're looking at, like, a lot of citizen scientists on OrcaGraph, you've got Orca names, you've got locations, you've got all sorts of things.

What you don't have is... I found Amazon was really good at picking out potential products, pricing, and sometimes location. It makes sense, right? Like... what they're good at, fabulous at picking those out. I don't need any of those things. So that was a very convergent data set, very few overlapping there.

Google was amazing at location, which was very helpful. It was pretty good at getting people's names. Which was really nice, and then it was also good at just getting any noun on the planet that was not "orca", or "orca name", or "orca sighting".

So I got a lot of, like, pings on "ice cream". And I was like, thank you, Google. Maybe not. It was an interesting lesson in, okay, yeah, no, you really do have to sometimes fine tune these, especially if you're going to work on some funny graphs, like some funny data sets that aren't as... Turns out, orcas and their indigenous names here are not as common in the vernacular and are not trained into these corporate models.

Shocking. But I think the bigger story isn't so much that, like, these didn't work in those, but that, like, having some way to easily identify that, not only makes analysis easier, but it also means that my graph was suddenly cluttered with ice cream, and I could suddenly remove, I could easily go back, remove it, go back to what I had already tried, and not be afraid to experiment.

And I think that's the big one, is like, finding ways to make it not scary to work in your knowledge, like, to work in Neo4j, to play with your database. Because I think that's where the biggest learning and the biggest insights will happen.

[00:11:26] Andreas Kollegger: Unless you're trying to come up with a Ben & Jerry's flavor. And then...

[00:11:29] Melly Beechwood: And then! Google's your friend.

[00:11:32] Andreas Kollegger: Seaweed, something, pistachio. I don't know.

[00:11:36] Alison Cossette: The Orca inspired line from Ben & Jerry's.

[00:11:39] Melly Beechwood: Yes. It is interesting. I was like, wow, I had no idea so many people were eating ice cream while seeing orcas.

[00:11:51] Alison Cossette: You never know what relationship is going to pop up once you start using knowledge graphs, Melly.

[00:11:55] Melly Beechwood: Pretty funny.

[00:11:56] Alison Cossette: I want to shift a little bit and I want to talk about talks really quickly. So have you done a lot of conference talks before?

[00:12:06] Melly Beechwood: Oh my gosh, no. So this is my second one ever. So my first one was also on this topic, but more of a general overview, with the Puget Sound Women in Data Science Conference earlier this year.

[00:12:20] Alison Cossette: What inspired you to apply for NODES?

[00:12:25] Melly Beechwood: You know, frankly, I actually didn't think I was going to get in. I was like, Oh, you know, it's a big conference. This is only my second time applying. And it was actually really late at night. I was looking at the conference lists and I was like, you know what, I should just do it.

And, sneak tip here. I totally had ChatGPT help me write the description because it was super late at night. I think I'd had a beer, you know, nothing too crazy. But, I was like, had a little bit more bravery, knew I wasn't gonna do my best writing, but also knew I probably wouldn't have the bravery the next day. And really used it as an editor to be like this is all my stuff. Can you make that sound... normal?

[00:13:11] Alison Cossette: This is the world that we're living in, right? I mean, that's, what's so wonderful about the innovation of these tools, right? So whether it's because it's late and you're snuggling up with your IPA or whether it's because, you know, you're a student who has a writing issue. The fact that we have tools to help us with that scaffolding I think is great.

[00:13:31] Melly Beechwood: I absolutely love it.

[00:13:33] Alison Cossette: Honestly, I'm really glad that you had both of those things because, obviously, we were super happy about your talk and we're really looking forward to it.

[00:13:40] Melly Beechwood: It's been a big confidence boost and it's a good reminder to everybody else out like just because it's something you're working on doesn't feel that big to you or it's just like your passion project doesn't mean other people aren't interested in it.

And it's good to be brave. Go out and do it. You never know where it ends up.

[00:13:58] Alison Cossette: Oh, that's going to be the pull quote of the day. "It's good to be brave!" By Melly Beechwood. Jenn, ABK, any other kind of, like, questions or lingering things that we want to, you're curious about from our special guest?

[00:14:17] Jennifer Reif: I was wondering, since this is only your second time, what are your takeaways from it? How did the first round go, you know, without hopefully putting you too much on the spot?

[00:14:27] Melly Beechwood: No, it's fine. First round has been going well.

Like, it was well received the first time around. This time around, I changed it up a little bit instead of just being about what I did, into focusing a little bit more on what people could do similar. And I really did that around the idea of not only is this, I think, something that could get a little bit more impact.

Personally, I would love it if more people could go out and help some endangered creatures out there. But also it's a bigger audience and I really want it to be something. So that's a bit of a stretch for me where I'm trying to, like, look through the talk and figure out.

I definitely think it's nerve-wracking. Running it through with some friends, or different things like that, can really help. We'll see. Hopefully it goes okay.

[00:15:18] Jennifer Reif: I'm sure. One of the things I really loved about your session was, it was something that was kind of unusual and you could kind of see that passion project coming through. And that's the kind of thing that we enjoy seeing as well.

And I had somebody else tell me that actually, too, is find something that you're really interested in. And chances are that somebody else will be interested in it. And even if they're not, necessarily focused around your topic, they'll see the passion that you have for that and go out and search for their own as well.

[00:15:48] Melly Beechwood: That's so true. I really did not think so many people would be as interested in the orcas as I have been, and that's really the heart of it, is there, the passion kind of draws people in and lets them see what they could do and really ignite that.

And that's such an honor to be on this side of it. I've been to so many conferences and been on the other side, and it's such an honor to be able to do that and get this opportunity.

[00:16:14] Andreas Kollegger: I just want to comment here that I love everything you've been saying about how you've been approaching this and agree with all the comments about like, you know, if it's interesting to you for sure, it's interesting to at least one other person, probably 10, probably a hundred. I don't even know about it, or even more than that. Right?

But you'll find friendship, you know, in this and that you've taken it up the next step and started to think, okay, not only is this topic interesting, but this is impactful. And how can I take what I've learned and how the impact is and like help inspire other people to actually have that same kind of impact?

I think all of this comes together. Inspire people to take that brave step and you will find good reception and actually you will have impact and it's, I think it's fantastic. I love the approach you're taking to this, so we are lucky to have you at the conference, honestly is what the situation is.

[00:17:01] Melly Beechwood: Aw, thank you so much.

[00:17:04] Alison Cossette: We are. Speaking of the conference, for those of you who don't know, you'll be able to see Melly at our virtual online conference called NODES this coming October 25th later this month for this episode. We have I think we run a full 24 hours, right, ABK. Because we've got an APAC series, an EMEA series, and Americas series as well.

So wherever you are and whenever you are, there will be something for you, but we obviously suggest you don't miss Melly.

[00:17:32] Andreas Kollegger: The sun will not stop shining on graphs.

[00:17:36] Alison Cossette: For 24 hours, we'll be in the sunlight. One of the things that we do here every month, Melly is on our podcast is we talk about our tool of the month.

And I was wondering if you had something that you've been using lately that really jumps out for you that you want to share with folks.

[00:17:54] Melly Beechwood: Yeah, so probably the best one for me right now that I'm loving is the Reflect app. It is a note taking app that's not, like, it's kind of in the same line of like Obsidian or Craft or some of those if you've heard of them, Notion even.

But it's pretty pared down where it's mostly just markdown. Really easy to use. Really, I think it's really pretty. Very fast, like lightning fast. Works everywhere. And then my favorite is it does have the graph map, so you can see all of your thoughts and them together. I know that's like really kind of controversial.

Most people find that useless, but I, maybe it's because we're here at NODES. I love it. I think it's so fun to see, like, all the different from a bird's eye view where you can see like, oh, hey, look, there's a section there. Maybe I should write an article about that. Clearly, I've done a lot of thinking there.

[00:18:50] Alison Cossette: No, it's true. I mean, yeah, I mean, to your point, it's this idea of what we learn from graphs in general is we see those communities of, in this case, of ideas. Or we see different kinds of clustering. And so much of what I find interesting about graph is just the structure itself that is created.

Once it builds organically, you can see where those different pieces have density, as you say. So I love that. I think it's a great idea.

[00:19:22] Jennifer Reif: I think it really brings home, too, the fact that we already think in graphs. We just don't necessarily realize it. And so seeing something in like a note taking app where you can kind of see exactly how your brain is connecting different things and pulling together things I think is super useful for, kind of, bringing that home.

[00:19:41] Melly Beechwood: I agree. It would be cool if at some point someone does one of these notes app in like a VR experience so you can like, kind of Iron Man style, like walk into it and play with your thoughts. I don't know, that would be cool.

[00:19:57] Alison Cossette: You continue to inspire, Melly, everywhere you go. Good ideas.

[00:20:01] Melly Beechwood: I don't do VR, so someone else is going to have to take that one on.

[00:20:07] Alison Cossette: Listen, when you inspire, you don't have to do it yourself.

[00:20:09] Melly Beechwood: Exactly.

[00:20:12] Andreas Kollegger: Throw it up to the audience, somebody out there will get some funding and, you know. They can thank us later.

[00:20:18] Melly Beechwood: Beautiful. Exactly.

[00:20:21] Alison Cossette: Exactly. ABK, what's on your list for Tool of the Month?

[00:20:27] Andreas Kollegger: Well, it's always hard for me to say anything other than Arrows, and I will be saying Arrows again because it's the app that I think about and use the most.

And I've been recently, so just this week, it's pertinent to me because I've been at this academic symposium that's all about graph drawing. So a lot of papers, you know, academic papers about, you know, all of these very interesting aspects of graph drawing. And it's re inspired me to get back into the code of Arrows and like, see what things I can tweak and what I can kind of apply from what I've learned this week.

So I unavoidably have Arrows on the brain yet again. I promise in a future month, I will learn something else that's a good tool. Gosh, it's just so good.

[00:21:08] Alison Cossette: I use it all the time. I'm working on... Jason and I are heading out to PyCon India next week. And we've, I always use it whenever I'm doing slides to really do different kinds of diagrams and explanations.

So I use it all the time too. So I'm going to, I'm going to thumbs up your tool of the month for sure. How about you, Jenn? What's on your, what's in your toolbox this month?

[00:21:32] Jennifer Reif: Okay, so I'm taking a little bit of a sidestep here. I kind of explored something new this month, looking at JBang, which does basically scripting for Java.

So I wanted to switch those words around, Java scripting, and that is totally something else. Too close to JavaScript. This is not JavaScript. This is scripting in Java. But yeah, JBang is really cool. I've heard a lot about it. I had never taken the opportunity to play with it. And I was working with some APIs that I was trying to import data into Neo4j.

And one API, you know, worked just fine, you know, just pulling it in... my automatic is to go to APOC utility library with Neo4j and immediately pull that in through browser, you know, hit the API, pull in data, you know, run Cypher, that sort of thing. But one of the APIs I was hitting required headers on the requests.

And the only way to do that is one particular APOC procedure. And the APOC procedure is blocked in Neo4j Aura. So I was like, okay, that's not gonna work. I gotta, I gotta do something else. So I had been recommended that I try some kind of scripting. And since I do a lot of stuff in Java. I'm like, hey, this is a great excuse to look at JBang.

And so I started digging into that and a little bit of a, of a learning curve. I need to do a little bit more research on some things because I can pull in dependencies. But it doesn't automatically do class imports for me. So when I try to like use a particular class, it says, "hey, you can't find this class".

So a little bit of a learning curve for me, just because I'm so used to using, you know, Maven and automatic, you know, stuff that it's just the IDE will just pull in all that stuff for me.

And this doesn't necessarily, unless I can find, unless I can find a really cool plugin that will do that for me. And I haven't come across that yet. So that's been kind of the thing I've been exploring. It's been really fun. Just a, a couple little gotcha things that I'm, I'm still figuring out seeing if there's an easier way to do. But yeah, really cool tool.

[00:23:31] Andreas Kollegger: Jenn, can I ask you? Could you, could you place it in kind of like the Java, I guess, tool belt, like, you know, you can do full on Java. There's other JVM languages, of course, over on, I'll say, this side with my hands disappearing. Imagine my hands were there. And then, of course, there's Java. And then there's Groovy, which is, okay, it's a separate language, but it's kind of on the scripting side.

[00:23:53] Jennifer Reif: Okay.

[00:23:54] Andreas Kollegger: Is JBang kind of ish groovy or like, what's, what's the experience like?

[00:24:00] Jennifer Reif: So I'm not super familiar with Groovy. But my immediate tendency when the Aura procedure wouldn't work, or the APOC procedure wouldn't work in Aura, was I could do something like a Linux script. And basically make like curl commands, pull the data, dump it to a file, that sort of thing, which works, but you're usually surrounding your commands in quotes and it's a little bit messy and sometimes a little bit more complex than what you need it to be.

And so, jBang allowed me to use Java logic, Java constructs, all of the if-then-else, and the exception catching and all that sort of stuff that I'm used to in Java. But, use it as more of a scripting sort of format where it's very condensed. You run it just in, it's usually like a single file, you run it from a command line like you do a Linux script. It just makes things really easy without having to go to the whole, I guess, process of creating a whole application and JAR-ing it and all that kind of stuff.

[00:25:15] Andreas Kollegger: So it's total interpreted runtime. You don't have to compile, you know, all that nonsense.

[00:25:19] Jennifer Reif: Yep. Yep. Yep. Pre compiled. It does that when you execute the Java file.

[00:25:27] Andreas Kollegger: Cool. Sounds good.

[00:25:29] Jennifer Reif: It's a lot of fun.

[00:25:31] Alison Cossette: Nice, so my tool of the month is actually a Streamlit app that was built by one of the folks here at Neo4j, Dan Bukowski, who I work a lot with around graph data science, and he built something called Agent Neo. And really what it does is it answers, it's a chatbot, but it will answer your specific graph data science questions and questions you have about Neo4j.

So if you're looking at trying to understand different kinds of algorithms, if you're curious about which centrality algorithm is best for you, or you're having a challenge with something, Agent Neo is great because you can talk through whatever your challenge is within GDS. So, as you know, at Neo4j, we love producing content and we love answering your questions.

And so Dan went through and took questions from community, documentation, different examples, and really pulled all of that together within the GDS realm to help answer those questions for you. He also has a blog associated with it that we'll have links to in the show notes, where you can actually see how he did it and what that looks like.

So that's my shout out for today is to Dan Bukowski and Agent Neo.

[00:26:47] Melly Beechwood: That's so cool. I'll definitely be checking that one out.

[00:26:51] Alison Cossette: I know. I love that we work with such innovative, interesting people, right?

[00:26:55] Jennifer Reif: There's always something new to explore. Now I have two. I use Arrows already, ABK. But now I need to check out Reflect, and I need to check out Agent Neo.

[00:27:05] Alison Cossette: Yeah, absolutely.

We do have a lot going on at Neo4j. Recently, we had a community project called Neo4Cyclone. And Cyclone DX is a SPOM or software bill of materials standard that's meant to reduce cyber attack risk. And the app can actually parse and ingest data with the standard into a Neo4j instance. So we're always interested in seeing, as we said, what is the community coming up with these days?

We're going to move on to some articles that have been coming up recently. We've got the Neomodel, which you may or may not be familiar with. It's the Python OGM for Neo4j. And what it does is it extends version support. And so it's going to allow the OGM community library for Python and Django.

And it's the open source community driven package. So we definitely have that that's available for you. And then we get into the things that are always super interesting for me, which is, path finding. So there's recently been something that came out specifically leveraging regex for Cypher. And so in this article, it takes us from Denmark Hill to Gatwick airport with quantified path patterns.

So whenever we're working on pathfinding, there's lots of different ways that we can weight different options. So whether we're talking about different weights that are either positive or sometimes you might want to use a negative weight. We really want to look at this. So in this case, we're actually looking at quantified path patterns. So that can be interesting.

We at Neo4j have a lot going on as far as our text goes around LLM. So there's a number of different articles this month and I'm going to group them all together.

The first one is around leveraging LLMs for graph data science pipelines. So anyone who's been working with GDS and has been working with those pipelines, you might have been wondering how that all works together. In this, there's a six step plan for how you leverage those large language models and four steps to avoid the pitfalls of ChatGPT. Talk about how to engineer prompts, how to test and analyze results.

There's a lot of work going on these days around understanding the results of what's coming out of the LLMs. A lot of interesting things around GenAI, MLOps. So you might want to start there.

We also have a couple of things around LangChain. So for people who don't know what LangChain is, LangChain is an open source library that specifically allows you to work with different aspects of LLMs.

It helps you do different kinds of integrations, different connections. And recently, we have some folks here who have been contributing to that open source library. So as of now, thank you, Tomaz Bratanic, the LangChain library now adds full support for Neo4j's vector index.

That's my favorite product improvement of the month is the release of vector indexing within Neo4j. So, LangChain Vector now includes a wrapper for Neo4j's vector search and in this particular article, Tomaz actually walks you through the high level process of using LangChain in this case with OpenAI, and it allows you to create, say, a Q&A system on Wikipedia data.

Again, similar to what is happening with Agent Neo. Also with LangChain and OpenAI and the vector space, there's a longer form version of this article, which is much more detailed and actually shows you how to set up and build your grounding demo with the actual vector embeddings. And this was done at a hackathon.

So, we definitely suggest that you check it out, really get interested in how you can use vectorizing. You know, similar to almost what Melly was talking about when she was talking about leveraging your own data to understand what you're working with. You know, it's going to show you different kinds of clusters. So when you're working with an LLM, it allows you to not necessarily tap all of the data that's in your knowledge graph, but to really understand which space you want to get to.

And of course, we've got some really interesting things around how we're creating knowledge graphs, leveraging LLMs to create knowledge graphs from unstructured text.

So our coworker, Noah Meyerhoff, has done a video as well as an article on how you can actually construct knowledge graphs from unstructured data. And what it goes to, Melly also referred to this earlier when we're looking at entity doing entity recognition. So it will go through, it will create the entities, and then once those entities are extracted, then it will create the relationships among them.

So whether you're into reading articles or you want to see the video, both of those are available and in the show notes. And then the last one I'm going to talk about really quickly is the Azure OpenAI Neo4j demo. So we've got some demos that are available with VertexAI from Google, but we also have a new one that's available where we look at Neo4j, OpenAI, and Azure, and so that will also show you how to create a knowledge graph from different types of data sources, so that's available in our videos as well.

Anybody else have any videos or articles that are striking you these days?

[00:33:00] Jennifer Reif: I will pipe in a little bit. There was an article out there for graphs for DFIR, which is digital forensics and incident response, which I think is pretty cool. We don't have a ton of content in this area, and so seeing something like this pop up is kind of fun.

The author talks about getting a holistic view of the network, helping to understand the structure of a cyber attack, then helping them detect anomalies at a very large scale and looking at suspicious access pathways or lateral movements with that. So, covers kind of the digital forensics incident response of cybersecurity goals using graph, and just a really neat article. So that'll be linked down, too, for anyone interested in looking at that.

And then I also wrote a blog post this month on how to verify a database connection from a Spring Boot application. So I was working on a project for a new Graph Academy self paced learning course. So, that's a very subtle promo for that coming up, hopefully in the near future.

But I was trying to find a way... I want users to be able to create the application and then I just want them to test the database connection. Just to make sure they have that part set up, before they go into adding the data model and creating queries, returning results from the database, and all that kind of stuff.

So, come to find out, there is a verifyConnectivity method that you can use in order to test that. And so it kind of goes through my process of how I learned to use this and what the best approach for my use case. Now, there's lots of other options. There's much more robust ways to go about testing your database and all that.

So that's kind of my article on that. That's available and will be linked down below as well. And then last but not least, there was a video by Sebastian Daschner talking about building applications with graph databases, specifically Neo4j, and he used Quarkus.

So Quarkus is a, Kubernetes native Java stack that helps you create Java applications that are kind of cloud native in nature. And so he uses Quarkus and then OGM, Neo4j OGM, which is Object Graph Mapper. So that helps him map the entities in the database to his Java objects. And so he goes through building a coffee flavor application and then a recommendation engine at the end. So that's a video there, walks through everything.

A lot of it is live code. So if you want to see kind of that kind of stuff, definitely check that video out.

[00:35:32] Alison Cossette: As we were having a conversation pre recording today, talking about how busy fall conference season is and fall events. So there's lots and lots of events coming up. The beginning of the month, we have QCon in San Francisco. There's also GraphTalk in Milan and we have a virtual Road to NODES - Neo4j GDS with GenAI on October 5th that I will be hosting with Dan Bukowski that we mentioned earlier.

Also we will be at DockerCon out in Los Angeles that first week of October as well. Moving into the middle of October, we have GraphSummit Frankfurt, we have another Road to NODES virtual about analyzing the physical world. There is also a QGIS Plugins with Python coming up, and Airplane Route Optimization using Neo4j, all happening on October 11th.

October 15th, we have ATO 2023 and "Building Open Source GIS plugins with QGIS, Python and Neo4j". Bringing us towards the end of the month, we have FOSS4G and we'll be "Building Open Source GIS plugins with QGIS, Python, and Neo4j". That's going to be in Baltimore. We're sensing a theme, aren't we? We also have GraphSummit for Government in Arlington, Virginia.

And... The big thing coming up, as we've mentioned many times, October 25th, NODES 2023. Rounding out later that week, we've got the Madrid Tech Show October 30th, as well as ODSC West out in San Francisco, where I will also be. And if you want to sign up for NODES, we do have the registration in the show notes.

Registration is free. It is virtual. It is 24 hours. We would love to see you there. And you certainly cannot miss the lovely Melly Beechwood at her first NODES presentation, where she will be showing us about graphs in the wild and for the wild.

Melly, thank you so much for being with us today. Was there any other sort of conferences that are coming up for you that you're going to be attending or you're looking forward to?

[00:37:53] Melly Beechwood: You know, my end of the year is pretty busy with Axon, so I am taking a step back after NODES but then I am excited to check out what's coming up in the spring and see both what to attend and what I might be speaking at.

[00:38:11] Alison Cossette: Alrighty, well that's going to wrap us up for the October edition of GraphStuff.FM. Jennifer, ABK, and Melly, thank you so much all for participating. And we look forward to seeing everyone in the future. And definitely at NODES 2023 later this month.

[00:38:31] Andreas Kollegger: Stay connected, people. Cheers.