Automatic first cut processing of Transcripts
Budget: £250 – £750 GBP
I have 590 YouTube Transcripts and 440 of them relate to videos included on the WordPress blog, and are of the most interest. The subject matter is health related, so nutrition, recipes, exercise and gear.
I know that NLTK can be used to split text into paragraphs
https://stackoverflow.com/questions/43193018/how-to-split-text-into-paragraphs-using-nltk-nltk-tokenize-texttiling
There are blocks of the transcripts that are a pitch for a product bundle that's currently on offer. The word bundle is mentioned with reasonable density, and the price, which I believe is $49. I am buying human hours currently to remove this type of text. I would imagine an algo that could analyze something like sentence by sentence, and remove the contiguous sentences that seem to be about the bundle, until there are 5 in a row not related to the bundle, in which case the scan can start again, because there's often another block of the same near the end of the transcript. I want to remove these paragraphs or flag them.
I know that NLTK can be used to split text into paragraphs
https://stackoverflow.com/questions/43193018/how-to-split-text-into-paragraphs-using-nltk-nltk-tokenize-texttiling
There are blocks of the transcripts that are a pitch for a product bundle that's currently on offer. The word bundle is mentioned with reasonable density, and the price, which I believe is $49. I am buying human hours currently to remove this type of text. I would imagine an algo that could analyze something like sentence by sentence, and remove the contiguous sentences that seem to be about the bundle, until there are 5 in a row not related to the bundle, in which case the scan can start again, because there's often another block of the same near the end of the transcript. I want to remove these paragraphs or flag them.