Hey #ozhist #twitterstorians, if you want to explore the gaps & inconsistencies in @TroveAustralia’s coverage of digitised newspapers the #GLAMWorkbench can help! Here’s a few suggestions...
@TroveAustralia For the big picture, see this notebook which visualises the total number of newspaper articles by year, then breaks them down by state. nbviewer.jupyter.org/github/GLAM-Wo… 
@TroveAustralia Notice the copyright cliff of death in the charts above? If you want to find out what newspapers are available beyond 1954, you can grab a list here: nbviewer.jupyter.org/github/GLAM-Wo…
@TroveAustralia But perhaps you’d like to compare the number of articles over time by newspaper. You can do that by using the QueryPic notebook. Go here and click on the 'Run live on Binder in Appmode’ link. glam-workbench.github.io/trove-newspape…
@TroveAustralia The notebook loads as a simple web app. Get a Trove API key, pop it in the box, and you’re ready to start.
To compare coverage between newspapers, click the ‘Compare newspaper’ tab, load the list of titles, then select your newspapers. Use a single space as your search term.
To compare coverage between newspapers, click the ‘Compare newspaper’ tab, load the list of titles, then select your newspapers. Use a single space as your search term.
@TroveAustralia You can also use the QueryPic notebook to examine shifts in language, and how the results from particular keywords changes over time. Just click on the ‘Compare queries’ tab and add as many different keyword combos as you want.
@TroveAustralia Note that you can easily switch between viewing the raw number of results per year and the proportion of matching articles per year.
@TroveAustralia You can also download your results as an image, an html page, or just grab the data as a CSV file.
@TroveAustralia If you want to dig deeper, Trove’s facets let you slice and dice the complete newspaper dataset in a variety of ways. This notebook goes into the options in great depth. nbviewer.jupyter.org/github/GLAM-Wo…
@TroveAustralia If you’re wondering what effect OCR correction has on your search results, you can do some analysis using this notebook. nbviewer.jupyter.org/github/GLAM-Wo…
@TroveAustralia There’s a range of other tools and examples in the Trove newspaper section of the #GLAMWorkbench. glam-workbench.github.io/trove-newspape…
@TroveAustralia But what if you find some interesting anomaly using QueryPic and you want to examine it in detail? That’s where the Trove Newspaper & Gazette Harvester is useful. glam-workbench.github.io/trove-harveste…
@TroveAustralia The Harvester lets you download metadata, full text, & even images from thousands of newspaper or gazette articles. Remember that the Trove web interface only shows you the first 2,000 results. Using the harvester you go much deeper into your results set.
@TroveAustralia There’s an easy-to-use, web app version of the Harvester. Again, click on the ‘Run live’ link to get started. Then you just plug in your Trove API key and the url of your search from the web interface, and the harvester will create your dataset! glam-workbench.github.io/trove-harveste…
@TroveAustralia Of course there are many other problems to confront & questions to raise as historians making use of large digitised collections like Trove. The recent Trove update, for example, has made it difficult to exclude tags from your search queries — what’s the impact of that?
@TroveAustralia But if we’re really want to examine these questions in depth we need to take search interfaces seriously as sites of critical analysis. I explore this a bit more in my article ‘Hacking heritage: Understanding the limits of online access’: timsherratt.org/blog/hacking-h…
So given the methodological implications of the use of @TroveAustralia & other online collections for #ozhist, where is the digital training in undergrad/postgrad courses?
