Simple book categorization based on book summary

Prompting
1

I would like to use OpenAI Embeddings API to categorize various media based on their titles and summaries as given by WorldCat. Categories are as follows:

  • Books
    • Action/Adventure
    • Art
    • Biographical
    • Business
    • Comics/Graphic novels
    • Computer/Internet
    • Crafts
    • Crime/Thriller
    • Fantasy
    • Food
    • General Fiction
    • General Non-Fiction
    • Historical Fiction
    • History
    • Home/Garden
    • Horror
    • Humor
    • Illusion/Magic
    • Instructional
    • Juvenile
    • Language
    • Literary Classics
    • Magazines/Newspapers
    • Math/Science/Tech
    • Medical
    • Mixed Collections
    • Mystery
    • Nature
    • Philosophy
    • Pol/Soc/Relig
    • Recreation
    • Romance
    • Science Fiction
    • Self-Help
    • Travel/Adventure
    • True Crime
    • Urban Fantasy
    • Western
    • Young Adult
    • Guitar/Bass Tabs
  • Music
    • Individual Sheet
    • Individual Sheet MP3
    • Instructional Book with Video
    • Instructional Media
    • LTP/Jam With, Lick Library
    • Techniques/QL, Lick Library
    • Complete Editions, Music
    • Music Book
    • Music Book MP3
    • Sheet Collection
    • Sheet Collection MP3
  • Radio
    • Comedy
    • Drama
    • Factual/Documentary
    • Reading

For example, the title (orange dashed) and summary (red dashed) below:

image
image856Ɨ318 75.9 KB

I am just using curl at the moment to get a hang of how to work with the API, but I am ultimately comfortable using any programming language.

Thank you for any assistance.

2

I believe I came up with a solution.

I create a loop that compares the categories with the target text one category at a time. Then I calculate the cosine similarities between each category and the target text. The closest cosine similarity will identify the closest category label.

Iā€™m going to try to implement this in JavaScript and let you know how it works.

1 Like
3

Nice. Thanks for stopping back by to let us know. Sorry nobody stepped up to help, but seems you arrived on your own. Thanks again.

1 Like