 Hi everyone. My name is Mark Westgate. I'm Science Advisor to the Atlas of Living Australia. I'm also a visiting fellow at the Australian National University, and I'm here today to talk about my R package. I've been working on it for the last couple of years or so. It's called RevTools. It's been about a year since the last update on Cranes, so this is really an opportunity to give you guys a bit of an update, so I hope you enjoy it. Just to give you an outline, I'm going to talk about what led me to develop RevTools. I'm going to talk about where we're up to now and what its current uses are, and I'm going to give a bit of a sneak preview of the next release, which I'm hoping will be out in the next month or so. So, for context, I put up this paper here, which I worked on back in 2015, and it came out of a situation where some colleagues and I were interested in looking at themes and a set of documents, and frankly just didn't want to go through a whole review process, honestly. We wanted to have a quick overview of what happens within that data set, and we found a whole body of literature in the computer sciences on all sorts of things, natural language processing, topic modeling, that sort of stuff that just hadn't really been used at that point in the field that we're in, which is ecology and conservation biology. And so we thought that was, you know, really neat, and we did some work on it, and we thought it was a useful insight to publish, and so we published this paper. That led, as part of writing that, that led to a code base, which ended up becoming RevTools. And so we published this last year in research synthesis methods, there's papers, you can see the DOI there, you can Google it as well if you want to find it. There's a preprint if they're charging you to get access to this from the journal. That describes what the software does, but I'm going to give you a quick overview here for those who don't know. As an aside, while I was developing that work, I realised that maybe some people don't want to use topic models, and it's a fairly simple thing to display some text using a shiny app. So sort of almost as an afterthought, I built in some manual screening tools and turned out these have been of interest to some people, which is great. So there's a screen abstracts version, which you can see here, there's one that's for screening titles only. And I think actually a bit more customisation of these would be desirable in future, which I'm hoping to get to at some point. So that's basically what this software does. And at the moment, to visualise that in a different way, here's a flow diagram. So basically there's a function for reading in your own bibliographic data. You can then pass that to a function that builds a document term matrix. So it literally counts the number of terms and which articles they occur in and plots those against one another. And then you can run a topic model. Now, I never wrote any topic modelling code, but a lot of the rest of this was inherently built within RevTools from scratch. You also have the option to, from importing your data with read bibliography, you can go straight to building your own topic models via MagDTM or you can pass your data to screen topics and it will do that stage for you. What's different about the next version that's nearing readiness for passing the CRAN is that we've outsourced our importing function or rather we spun them off. So synthesiser is a project that I worked on with Eliza Grains last year. It basically takes the RevTools import code and cleans it up a little and puts it in a separate package so that we can maintain that independently. That's been quite successful and quite happy with that as a piece of software. It's on CRAN already. What we've also done is moved to Quantita for building document term matrices. I trust their code more than I trust mine frankly and they do a really good job and so I'd suggest people check that out. And topic models, we've gone from the topic model package to structural topic models using STM which again is a very fantastic piece of software so I suggest people check it out. And the final thing that's changing is that screen topics will remain within RevTools of course but we're thinking of doing a full redesign just to make it a bit prettier than it used to be. The other thing to consider is that I've mentioned on the left that's the workflow that we currently have built into the package. That's fine but I noticed there are a few problems. One of which is that you can't pre-calculate a topic model and send it to screen topics which is annoying if you want to do your own parameterization of a topic model or if it's going to take a long time to run you don't want it to be running in a shiny app so now the new version supports you calculating a topic model however you want and then just passing it to screen topics as a visualization tool. Alternatively there's some new functions that just enable you to do plotting using ggplot of the sort of diagrams you get from screen topics but in a static way and that can be useful for people doing visualization for publications for example. The final thing that I'm adding is a function to preload a shiny app so basically what that means is that if you want to be able to control what someone else, if you're part of a team, you want to say this person should have 100 abstracts which are these ones in a particular order and a person too should have 300 abstracts and they should be in a different order. This function enables you to build a shiny app and store it as an RDS file and then you literally send it to the person however you want, Dropbox, email, whatever and they can load it with a single command simply the same as you would on your own machine screen abstracts and they'll load it without any option to reorder or change the settings. That can be useful also if you're doing things in experimental contexts like checking how long it takes people to screen things or the effect of doing things in a different order or with keyword highlighting for example. So in summary RevTools is a package for combined screening and topic model visualisation. I hope that this new set of tools make it more useful for a wider range of people and applications. The next round of revisions I'm hoping will be around enabling the user to more deeply customise their screening experience. I don't know when that's going to happen but I'm hoping still this year. So thanks for your time and look forward to hearing any questions.