0

Let's say I'm adding specific food product separated by commas.

Eg : The ingredients in tomato soup is tomatoes, salt, pepper. You need to heat the tomatoes till it smells like burnt apples.

Now by previous training it can detect 4 ingredients :

  1. tomato
  2. salt
  3. pepper
  4. apple.

Apple is the wrong span detected here.

But burnt apples is not an ingredient in the dish, it is just a random reference. SO how can I stop or reduce weight-age based on phrases or way the sentence is framed?

Also, I'm using spans (spancat). I've not used spangroup and acutally I don't know if spangroup will fix this issue.

Please let me know how I can add weightage for specific pattern or phrase in spans.

jason
  • 3,932
  • 11
  • 52
  • 123
  • If you want your model to distinguish ingredients from non-ingredient food mentions, you'll need to provide it with more training data, particularly lots of examples of non-ingredient food mentions (like apple). However that's very difficult to learn. – polm23 Oct 31 '22 at 04:09
  • Yes. exactly. Any work round? – jason Nov 02 '22 at 14:53

0 Answers0