 Hi everyone, my name is Sarah Young, I'm a librarian at Carnegie Mellon University and I'm really glad to be presenting some ideas about a new project that's just getting underway using network analysis and research weaving in evidence and gap maps. This work is in collaboration with Neil Hadaway, Shanichi Nakagawa and Max Callahan. So to start, I think many of you are familiar with evidence and gap maps as a method for bringing together scattered knowledge from existing research for policy and practice, and to highlight knowledge gaps and knowledge clusters. Evidence and gap maps typically include hundreds or even more than a thousand studies, and the findings are usually presented in thematic matrices of study characteristics as shown on the previous slide, and are analyzed across space and time. For example, this image shows a visualization of the geographic location of the studies included in the review, and you'll also often see analyses of the number or types of publications over time, for example. However, as Nakagawa and colleagues pointed out in their 2019 paper introducing this concept of research weaving, there are some real missed opportunities here with regard to what more these evidence and gap map data can tell us. And this is where network analysis can come in. So network analysis essentially adds an additional dimension of study to the evidence and gap map context, namely a social dimension in which we can begin to understand how this body of knowledge is structured through collaboration and citation behavior. And this could ultimately inform recommendations for future research, as well as aid in the discovery of additional relevant literature. So network analysis kind of simply put is a set of techniques that can be used to understand social structure and to visualize and analyze relationships among actors or objects. And so when we talk about network analysis in terms of research studies. This is sometimes also called science mapping. We're usually referring to one of several relationships upon which these networks are structured. So relationships between researchers based on shared authorship can be represented in what's called a co authorship network. And this type of network can help us identify, for example, key researchers or research silos, and can also indicate something about the flow of information and knowledge across the network. So basically the relationships between publications based on cited references, and this is typically done through co citation networks or bibliographic coupling networks. And here you see an image of a co citation co citation network. Also, you can look at bibliographic networks based on the occurrence of similar similar terminology. Key word co occurrence networks can indicate topic based communities. And these can allow for further analysis of collaboration and citation behavior based on topics derived using text mining models. And these can also show us new terminology that can be used to improve search and discovery of relevant research. So within the current work we're seeking to demonstrate the added value of network analysis for evidence and gap maps. And we're going to do this using open source tools and methods that can be applied widely in evidence and gap maps across various disciplines and domains. So we'll be using a previously published evidence and gap map on the role of vegetated strips and agricultural fields. Vegetated strips are intentionally installed to serve one of any number of rules. This could be mitigating pesticide runoff preventing erosion, increasing beneficial insects, just to name a few. And this particular evidence and gap map aim to map the evidence on the effect of these vegetated strips in serving these multifunctional roles. In addition to the sort of standard types of analysis conducted in evidence and gap maps. These authors also did some interesting text analysis and found a truly stunning array of terminology that is used to describe vegetated strips. And so this is something I think we can probe further using this network approach. There are different types of networks, the types that I just described, to provide a deeper level of analysis into the social structure of this research community. And there's a number of possible research questions that we may explore with this project. But ultimately, you know we really hope to provide sort of some understanding of how this research community is structured. And you know thinking about how this new information can inform the direction of future research on the roles of vegetated strips in agricultural fields. So there's a number of open source tools that are available to conduct network analysis and create network visualizations. I'm not going to be a fluffy and boss viewer to popular tools, but given the focus of this conference on our, I'll mention more specifically, the bibliometrics package. So most of the visualizations that you just saw in the previous slides were generated with bibliometrics. This package works with outputs from standard bibliographic databases like Scopus and Web of Science. You can use actually functions in the Litsearcher package, you're familiar with that, to build an exact title search from your list of evidence and gap map studies, and then run that search in the database to export a file that can be analyzed in bibliometrics. So that's sort of a nice workflow that can be implemented in R. I just mentioned a few challenges with these techniques. So for example, co authorship networks. It's important that author names be correctly disambiguated. And in a large data set this can be a pretty onerous task, but tools like open refine can be really helpful in this process. And we'd certainly welcome, you know ideas for other tools perhaps and are that facilitate this type of data cleaning. That would be true of author affiliation data. So if you wanted to understand institution level collaboration, author affiliation data can be quite messy to deal with as well. Another challenge is just getting the citation data for included studies to construct the co citation and coupling networks. So I mentioned the bibliometrics package in ours designed to work with outputs from databases like scopus and web of science. But of course in an evidence and gap map, not all studies, maybe index in these databases. And so this could involve potentially some manual work to add these on indexed studies to your bibliometrics data file. Thank you for being there. I'm happy to take questions via slack. Certainly also get in touch via email and definitely looking forward to your thoughts on this project. Thanks.