Error in C1M2 of RAG for computed TF-IDF scores?

Course Q&A Retrieval Augmented Generation
,
1

Hi,

I’m wondering if there’s not an error in the video of https://www.coursera.org/learn/retrieval-augmented-generation-rag/lecture/MYS0a/keyword-search-tf-idf at 6:05 with the IDF scores computed for each words based on the given word occurence in documents (see screenshot below).

For instance, for the word:

  • making, DF = 3/5 so IDF should be log(5/3) = 0.22,
  • pizza, DF = 2/5 so IDF should be log(5/2) = 0.4,
  • without, DF = 4/5 so IDF should be log(5/4) = 0.1,
  • a, DF = 4/5 so IDF should be the same as without log(5/4) = 0.1
  • oven, DF = 4/5 so IDF should be log(5/4) = 0.1

Could someone explain me what are the scores computed for each word if not IDF or log of IDF?

:thanks:

Capture d’écran 2026-02-07 à 13.53.45
Capture d’écran 2026-02-07 à 13.53.451920×1200 139 KB

2

hi @jdalbosc

yes you are right. Scores look mixed up.

for making df 1/6=0.1666

idf log(5/3)=0.22

tf-idf 0.1666 * 0.22= 0.03652 =0.037

@lucas.coutinho can you have a look at this issue.