← pandyaHomeLab

Quora Duplicate Questions

Bag-of-words + 22 handcrafted features · Random Forest · Quora Question Pairs

Two questions

Try an example:

Prediction failed. Please try again.

Prediction

P(duplicate)
0.50

Raise the threshold to flag fewer pairs as duplicates. That means fewer false "duplicate" calls (the costly error) but more real duplicates missed.

Why: the engineered features for this pair

Preprocessing pipeline

Question 1

Question 2

Model Card

Model
—
Train rows
—
Accuracy
—
F1 (duplicate)
—