Jeff Chen, Chief Data Scientist Star Ying, Data Scientist

Search String An alys is P r o j e ct Jeff Chen, Chief Data Scientist Star Ying, Data Scientist

Hopes and dreams go here

Go!

search query went here

Go!

Results for: “search query went here” 1. Ideally your best result 2. But just in case, here’s another 3. And another 4. And another 5. And another 6. We can do this all day 7. More to come 8. Here’s another one. 9. This one’s pretty good. 10. Below here, you probably won’t be paying attention 11. …I was wrong…

For search and retrieval to work… Text acquisition

Query parser

Text transformation

Synonym engine

Index creation

Ranking engine

+n

For search and retrieval to work… Text acquisition

Query parser

Text transformation

Space reserved for synonym engine

Index creation

Ranking engine

+n

Problem! Record is as good as the metadata and tags.

Problem!

W.I.T.A

Problem! Terminology evolves like Pokémon. (therefore it’s hard to catch match them all)

Problem!

[Pilot] Solution new ranking engine to recommend terms light tree

christmas

well

flow

holiday

wellhead

Solution logs Search logs 80 million search log records

Solution In Art Area i pr(tree | christmas) pr(christmas | tree)

Parallelizable python program • ETL of logs • Estimate ranked linear conditional probability estimates

Goal christmas christmas christmas christmas christmas

AND tree AND wellhead AND holiday AND light

Go!

Implications

8000+

600,000+

5

consistent

USPTO examiners

hrs/ application/ examiner

applications

office actions