Data Science Stack Exchange is a question and answer site for Data science professionals, Machine Learning specialists, and those interested in learning more about the field. It only takes a minute to sign up.
Sign up to join this community
I am looking for a machine learning model ideally with inference speeds of no longer than a few minutes that is able to take in n human written reviews and output a blog article from them.
The model would need to be pre-trained or if it does need training using the reviews then the training would not take more than a few minutes on a modern single GPU machine.
Can someone point me to such open-source projects?
There is a solution that could work well. It requires minimum effort but has to be tested.
If you take several reviews and you group all first paragraphs together, then the second ones, etc. and you apply an efficient summarization model, you should have the essence of all reviews.
The model would recognize the most frequent patterns in an organized way and do the job.
It could work with sentences instead of paragraphs.
If the articles have very different sizes, you can use summarization for each of them to 10 sentences, and then apply the process described above.
Note: You can't do this with full reviews next to each other because the model would not recognize the beginning from the end for each of them.
https://huggingface.co/facebook/bart-large-cnn
Other models:
https://huggingface.co/models?pipeline_tag=summarization&sort=downloads
