 Hi, and welcome to my talk. My name is Neil Hadaway, and today I'm going to talk to you about two apps that I've built for grey literature searching called GSscraper and greylitsearcher. And I'll hopefully show over the next 10 minutes that they are potentially useful but flawed tools for searching for studies and evidence synthesis. So before we get started, I just wanted to give a definition about grey literature, what it is. And it's been defined as any literature which is produced on all levels of government, academics, business and industry in print and electronic formats, but which isn't controlled by commercial publishers. So that's where publishing is not the primary activity of the producing body. And it's things like theses, organization reports, white papers and consultancy reports. And it's important for any review that wants to be rigorous because you're looking here for comprehensiveness and there's often a lot of really useful information that just wasn't aimed to be published in an academic audience. And it depends on your subject area obviously, but there are some systematic reviews where 25 or 30% of the evidence based in the final review is grey literature. And traditionally greylitsearching is done manually so people go to Google or other websites and search manually and only record when they find something relevant. So then we don't know how they went about searching perhaps. And even if they did report what they did, it's very difficult to then repeat it. So it's very laborious and largely unrepeatable. So I built a couple of tools to help. The first one is called GS scraper and the GitHub repository and shiny apt link are there on the top right. And the way that GS scraper works is to make use of the fact that Google Scholar creates a URL with patterned information that come from your search. So it's very easy to build a functional Google Scholar URL just using search terms. And that will then take you to a specific page with a specific search. And if you play around with a Google Scholar advanced search, you can see all other things that you can put in including authors, phrasing, whether to include citations or not. The journal is that you want to search in and years and stuff like that. And the way that GS scraper rather works is to first of all build all of these links based on the number of pages you want to search from and up to. It then saves these results and scrapes the code within the HTML files and then you can download it as a CSV or an RIS. And in the case of GS scraper, you can publish that search to Zenodo as a sightable search history. And the good thing is part of it works. So the first part where we're building functional URLs works really well still. And it actually does download the HTML files. The problem with GS scraper is that those HTML files are then held within the app. And the next part doesn't work, which is the part where it's scraping the code. So you can see the website with the Google Scholar search results on the left panel here. And then on the right, you can see the source code. And the problem is that the code that Google Scholar uses just fluctuates over time. And the problem is here that there's one specific code in the code that I've written that is using regular expressions to find the code GS underscore RI, which indicates each one of the 10 search results on the page. And that code has changed. So something within the website code has altered. The next tool that I built was called Gray Lit Searcher. And this is very similar to Google Scholar and GS scraper, but just on Google. And the way that this works is that it searches a given website. So let's say you want to search the Stockholm Environment Institute website for the phrase climate change, where you can search that site using Google Scholar site search. And it does the same thing. It builds the links. It downloads the HTMLs and then scrapes the search results as a CSV file in this case and lets you download the data. The first part works very much the same as GS scraper. It will download or it will produce functioning URLs based on your search terms for each website and for the number of pages of results that you want. And actually with Gray Lit Searcher, I've built in some extra code to let you download the HTMLs before it starts scraping them, which I didn't get around to doing in GS scraper. So you can transparently document your search, but you still then need to manually go through and screen everything. But again, the problem here is that the code that I've used to break up the Google search results is quite complex and it's easily changed. So this doesn't work as well. And so what I could do with both of these tools, and I was doing for a while, was to go through and unbreak them whenever the code changes, which requires quite a bit of time to play around and work out how it's changed. Unfortunately, I don't have that time anymore. I'm not an employed scientist. I've changed career. So I don't have time to fix the app. And I'm really sorry because I can tell that people are accessing it and trying to use it. And I know that's frustrating. A longer term solution to get around it is to avoid using regular expressions, which would mean that it might be able to use the structure on the web page rather than the underlying code, which is more likely to change in a very small way. And I don't know quite how to do that in a sort of future proof way. So I really need some help there. So yeah, if you are willing to help and have time perhaps to take over or to join a team, I'd be really happy to chat to you, perhaps even convene a hackathon over the next year maybe, or a hackathon at our next evidence synthesis hackathon event. If you are interested, you can drop me a DM, a direct message on Twitter, or you can raise an issue and just say that you're keen to take over. If you really want, you could actually start trying to fix it on GitHub by forking one of the repositories and having a go at fixing the code. I'd really, really appreciate that and that might be a really efficient way to do it. Or you can drop me an email. So if you're interested in a longer term hackathon and you've got some time, but perhaps don't have time to fix it yourself, maybe that's a solution. But yeah, thanks for your interest in the tools. I really appreciate it. I can see people trying to access it and I'm really sorry that they're not currently working, but I'm really trying with the resources and time that I have to find a longer term solution. Thanks again.