llama2_aided_tesseract
AI
ocr error correction
SHOW HN2023
How HN took it
0 positive, 4 neutral or asking, 0 mixed, 1 critical, out of 5 comments about the project. Thread score -0.17.
Performance: 0 for, 1 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Idea of LLM-based proofreading of OCR is promising
From 2 comments
I’ve been doing something similar recently that impressed me. I have been taking handwritten manuscripts from the early 1800s and feeding them into AWS Textract and then feeding…
AaronNewcomer on HNVery cool. I think this is an interesting benchmarking task for a language model (as well as the practical uses). I tried the same thing some time ago, just on a random snippet…
version_five on HN
Local LLM is free to run compared with paid APIs
From 2 comments
Cool. I'm sure that would result in better results, but then you have to pay per request and I'm sure it could get pretty expensive if you're talking about long documents. It's…
eigenvalue on HNI just realized that a very similar approach could also be very useful in correcting and reformatting automatic speech recognition transcripts. I made another tool recently for…
eigenvalue on HN
Objections
Llama2 hallucinates new sentences, needing filtering
From 2 comments
- The maker's post
I’d say that it does work pretty well. It could simply be that I’m sampling too much from the LLM which is causing it to hallucinate more than it should, hence why I needed to…
eigenvalue on HN
Hallucination risk makes it unsuitable without quality control
From 1 comment
I’d say that it does work pretty well. It could simply be that I’m sampling too much from the LLM which is causing it to hallucinate more than it should, hence why I needed to…
eigenvalue on HN
Extremely slow processing
From 1 comment
Requests
Try fine-tuning Llama2 on synthetic corrupted OCR text
From 2 comments
You should be able to build a text->image->corrupt with noise->tesseract pipeline to generate some synthetic supervised fine tuning examples (or just try corruption the text, I'm…
version_five on HNAs long as you can get tesseract to make similar text errors, the actual corruption doesn't matter though. You'd have to play with it but I'd guess the most important thing is to…
version_five on HN
Parallelise across pages with a Ray cluster
From 1 comment
Questions
Would different hyperparameters reduce hallucination?
From 1 comment
Is it possible that with different hyperparameters, LLAMA 2 would be more faithful to the original? I'm pretty surprised that it would hallucinate anything majorly different than…
Reubend on HN
Which Llama2 model was used and how well does it perform?
From 1 comment
Very cool. I think this is an interesting benchmarking task for a language model (as well as the practical uses). I tried the same thing some time ago, just on a random snippet…
version_five on HN
Alternatives named
AWS Textract with GPT-4 or Claude 2 correction
From 1 comment
I’ve been doing something similar recently that impressed me. I have been taking handwritten manuscripts from the early 1800s and feeding them into AWS Textract and then feeding…
AaronNewcomer on HN
ChatGPT for reformatting transcripts
From 1 comment
I just realized that a very similar approach could also be very useful in correcting and reformatting automatic speech recognition transcripts. I made another tool recently for…
eigenvalue on HN