I am a student in the Sequence Models and I came across a potential bug in the Week 2, Second Assignment (Emojify). I would like to report this issue for review and possible correction.
In the 1.5 - Examining Test Set Performance section, after Exercise 2, it is stated that:
“Because cherish has a similar embedding as love, the algorithm has generalized correctly even to a word it has never seen before.”
However, upon testing the algorithm, I found that the phrases ‘I love you’ and ‘I cherish you’ are actually assigned different emojis, indicating that the algorithm does not generalize correctly for the word ‘cherish’. This might cause confusion for students who are trying to understand how the algorithm works.
It would be helpful if this statement could be reviewed and clarified in the course materials to avoid any misunderstandings.
Thank you for your attention to this matter.
Best regards,
Reza Azami