47. Plan
● Process
● Examples
○ Recommendations
○ Advertisement
○ Duplicate detection
○ Fraud detection
○ Search
48.
49.
50. Q: Information
D1: Counting is not hip
D2: Information extraction is the process of ...
D3: Information gain is used to …
D4: Deep learning is better than counting
57. Cosine similarity + TF-IDF
We can use it for:
● Search
● Recommendations
● User profiling
● Near-duplicate detection
58.
59. Plan
● Process
● Examples
○ Recommendations
○ Advertisement
○ Duplicate detection
○ Fraud detection
○ Search
60. Key takeaway
● It’s okay to start without machine learning
● First, do heuristics (counts)
● Prove that it’s useful
● Iterate
61. mlbookcamp.com
● When you already have counting
● Time to add some machine learning!
● http://bit.ly/mlbookcamp
● Get 40% off with code “grigorevpc”
Machine Learning
Bookcamp