Large-scale NLP and AI copy for book reviews
Budget: £5,000 – £10,000 GBP
Hi there
I’m now starting a new project aggregating and summarizing reviews of books. It’s a huge project that depends on two main skills: (1) heavy scraping and (2) NLP and AI copy. This post is about NLP and AI copy. I have created a separate one for the scraping part as I figured we might need a separate person for that part.
Why this job is cool:
(1) it's a huge project = stable income, and you can combine with other freelance jobs as you like
(2) I have a profitable business (RunRepeat), which means that we would not run out of money.
(3) A lot of decision-making on your end, and zero bureaucracy.
It’s a big project, and you will be part of defining the direction of where we’ll go. Below are the bare fundamentals on where we want to get at. I understand that some of these require different skill sets, and you might not have them all. But, if you have fundamental skills in these areas and hands-on experience with something similar, I’d love to chat with you about this and see how we could work together. The tasks below are for both scrapings, NLP and AI copy - just to give you the overview of the entire project.
(a) Scrape Amazon and Goodreads for all book titles and store basic information about the book, author, categories etc. (Millions of books). Millions of books. Example page: https://www.amazon.com/Zero-to-One-audiobook/dp/B00M284NY2/ref=sr_1_1
(b) Find all critic reviews of that book title and consider how to match variations, e.g. some only mentioning the first part of the title, or another word for it.
(c) Do text analysis of each critic review to rate them 1-100 in how positive the critic is about the review itself. I understand that it will be hard/impossible to reach such a granular score, but maybe we’ll end up with a 1 to 4 rating scale like what this site has or similar: https://bookmarks.reviews/reviews/something-new-under-the-sun/
(d) Find the best way to make AI writeups summarizing critic opinions as well as book summary/introduction to have unique copy on our pages.
(e) Look for all “lists” where books are mentioned. For example “best business books” or “most recommended books to read in 2021” and suggest an algorithm how we incorporate these “buying guides” into our overall scoring system. How to weigh the first book listed, the 10th etc?
(f) Get an overview of all cases of where the book has been recommended by some person or entity. The end goal is that product pages would have sections like “Recommneded by: Elon Musk, Bill Gates” and then users can click on these tags to land on a page with all books recommended by Elon Musk. Store quotes.
(g) Scrape all user reviews and make text analysis to extract characteristics of the books.
(h) Get all awards for each book.
(i) Overview of all forum discussions, e.g. scraping of reddit and other sites that users can click on to read more. Provide short snippets.
(j) Based on user and critic reviews, get to an overall score, and create our own lists for all possible categories, like “best python programming books” or “most recommended” that can then be narrowed down with filters.
I know, it’s not something that you’ll have done by tomorrow. As a first step, let’s see if there’s a match between you, I and the project.
I’m now starting a new project aggregating and summarizing reviews of books. It’s a huge project that depends on two main skills: (1) heavy scraping and (2) NLP and AI copy. This post is about NLP and AI copy. I have created a separate one for the scraping part as I figured we might need a separate person for that part.
Why this job is cool:
(1) it's a huge project = stable income, and you can combine with other freelance jobs as you like
(2) I have a profitable business (RunRepeat), which means that we would not run out of money.
(3) A lot of decision-making on your end, and zero bureaucracy.
It’s a big project, and you will be part of defining the direction of where we’ll go. Below are the bare fundamentals on where we want to get at. I understand that some of these require different skill sets, and you might not have them all. But, if you have fundamental skills in these areas and hands-on experience with something similar, I’d love to chat with you about this and see how we could work together. The tasks below are for both scrapings, NLP and AI copy - just to give you the overview of the entire project.
(a) Scrape Amazon and Goodreads for all book titles and store basic information about the book, author, categories etc. (Millions of books). Millions of books. Example page: https://www.amazon.com/Zero-to-One-audiobook/dp/B00M284NY2/ref=sr_1_1
(b) Find all critic reviews of that book title and consider how to match variations, e.g. some only mentioning the first part of the title, or another word for it.
(c) Do text analysis of each critic review to rate them 1-100 in how positive the critic is about the review itself. I understand that it will be hard/impossible to reach such a granular score, but maybe we’ll end up with a 1 to 4 rating scale like what this site has or similar: https://bookmarks.reviews/reviews/something-new-under-the-sun/
(d) Find the best way to make AI writeups summarizing critic opinions as well as book summary/introduction to have unique copy on our pages.
(e) Look for all “lists” where books are mentioned. For example “best business books” or “most recommended books to read in 2021” and suggest an algorithm how we incorporate these “buying guides” into our overall scoring system. How to weigh the first book listed, the 10th etc?
(f) Get an overview of all cases of where the book has been recommended by some person or entity. The end goal is that product pages would have sections like “Recommneded by: Elon Musk, Bill Gates” and then users can click on these tags to land on a page with all books recommended by Elon Musk. Store quotes.
(g) Scrape all user reviews and make text analysis to extract characteristics of the books.
(h) Get all awards for each book.
(i) Overview of all forum discussions, e.g. scraping of reddit and other sites that users can click on to read more. Provide short snippets.
(j) Based on user and critic reviews, get to an overall score, and create our own lists for all possible categories, like “best python programming books” or “most recommended” that can then be narrowed down with filters.
I know, it’s not something that you’ll have done by tomorrow. As a first step, let’s see if there’s a match between you, I and the project.