Thread by @abobic_

So I finally hooked up CSX (my interactive graph-based vis. #analytics tool) to #OpenAlex (thx @OpenAlex_org for making this possible for free :))

You can checkout the app and explore the OpenAlex dataset as #interactive #graphs on:
Read more 🧵👇 https://t.co/F62UTt6JtEdemo.csxp.me

To demonstrate how someone might use #networkAnalysis I decide to explore the co-authorship network of Roche🇨🇭 I initially retrieved 200 papers from Roche 🇨🇭 by using the CSX advanced search and from there expanded the #network to 1k papers from Roche 🇨🇭 authors (see picture)

Next I identified the 4 most common collaboration institutions in these 1k papers and removed papers where none of them were present (right click on canvas and "remove not selected").

I then added papers that were co-authored by at lest one of these institutions and by Roche🇨🇭(right click on a node representing one of the institutions and select "context expand through node" which takes into consideration our initial query (roche) as well as the selected node)

The network you see here is constructed of purple nodes representing titles of papers and pink/red nodes representing the institutions. The institutions that co-authored a particular paper are connected to that paper. Next I looked at the concepts ...

Since I can dynamically change the network schema using CSX I decide to take a look at the concepts and the institutions that are most commonly co-appearing with particular concepts.

To find the most commonly appearing institution and concept pairs / groups I filtered the graph to only include edges with a weight of 20 or more (i.e. leave only concept and institution pairs that appear together in at least 20 papers) Obviously Roche is connected to all of them

I repeated the same exercise for the journals / conferences and institutions connections. Since these are only papers co-authored by Roche 🇨🇭 authors this can be interpreted also as "the types of research they collaborate on with particular institutions"

Finally I identified the most active author (42 papers) in this network. For the representation of their network I used a more complex schema that includes the author, papers, institutions, concepts and conferences / presentations.

During this whole process I had access to all the paper details and could have explored them through the "results panel" list view (you have to switch from table to list).

Since this is a research project there are at this stage some major performance issues lying around waiting to be solved :) So if you feel like suggesting some improvements feel free to jump to the project GitHub page:

GitHub:
Also I'm doing a usability study / experimenting with some "smarter" feature ideas so if you don't mind please also fill out the (anonymous) feedback survey and enable the (anonymous) tracking 🤓 Any feedback and questions are also welcome! 🥳github.com/aleksbobic/csx

Also forget to mention: built using #reactjs and #Python @FastAPI So if there are any 🧙‍♀️ 🧙‍♂️ 🧙 in one of these who think they might know a thing or two about speeding things up feel free to comment / ask / open issues :) (btw also trying to move from pandas to #polars