Data Journo at Indian Express Online • Previous: Hindustan Times, IndiaSpend • Alum: WashU, St Louis & Hindu, Delhi • mail@shijith.com

New Delhi, India
Joined January 2008
Doing a project on ncert history books & how they've changed over time. Building in public on this so people can see how the work is being done. Done 2 threads on it so far: 1st thread -- xitter.cf/shijith/status/2102963… 2nd thread -- xitter.cf/shijith/status/2105926… Starting the next one now👍
So i'm doing this "build in public" thing where i talk openly about a project i'm working on. This current one is about ncert history textbooks & how they've changed over the years. Already done one thread about it (link below), starting the next one now👍 xitter.cf/shijith/status/2102963…
1
162
the 'judging' was done during benchmarking at the link below. In the end, the models on top of OmniDocBench were on top in our benchmark too. But a model, GLM-OCR, that was 5th in OmniDocBench was last in our benchmark. Make of that what you will! huggingface.co/spaces/shijit…
2
46
Ending this thread for now. Will be back later in the week to talk about the frontend for this project, what the user interface will be etc. 👋
9
So i'm doing this "build in public" thing where i talk openly about a project i'm working on. This current one is about ncert history textbooks & how they've changed over the years. Already done one thread about it (link below), starting the next one now👍 xitter.cf/shijith/status/2102963…
Am going to do something a little different over the next few weeks. Have started on a new project, and I'm going to "build in public". So this is something people in tech do a lot, where they post about their product, as its being built. So they do it partly for transparency...
2
1
308
remember the rest of their lives. And what they tend to fall back on as adults, influencing their views, opinions, votes etc. But even then, there's still value in seeing how a student's first exposure to the humanities in 6th-8th has been regulated under different regimes...
1
9
I was going to talk further about how i extracted text from the pdf files, but I see this thread has already gone on long enough, so I'll stop now. Will do a separate thread on Sun/Mon that goes into the technical stuff. See you then! 👋
9