Architecture of the Triposo travel guide Douwe Osinga (@dosinga) - - PowerPoint PPT Presentation

▶

Nov 08, 2022 130 likes •821 views

Architecture of the Triposo travel guide Douwe Osinga (@dosinga) Thursday, 17 October 13 The Company Thursday, 17 October 13 Founded by Ex Googlers Thursday, 17 October 13 In Sydney Thursday, 17 October 13 Financed in California Thursday,

SLIDE 1

Architecture of the Triposo travel guide

Douwe Osinga (@dosinga)

Thursday, 17 October 13

SLIDE 2

The Company

Thursday, 17 October 13

SLIDE 3

Founded by Ex Googlers

Thursday, 17 October 13

SLIDE 4

In Sydney

Thursday, 17 October 13

SLIDE 5

Financed in California

Thursday, 17 October 13

SLIDE 6

Headquartered in Berlin

Thursday, 17 October 13

SLIDE 7

Distributed Team

Thursday, 17 October 13

SLIDE 8

Quarterly Jamborees

Thursday, 17 October 13

SLIDE 9

Last one on a bus in Spain

Thursday, 17 October 13

SLIDE 10

Algorithms are King

Thursday, 17 October 13

SLIDE 11

No Human Editing in our Guides

Thursday, 17 October 13

SLIDE 12

The Product

Thursday, 17 October 13

SLIDE 13

Our Mission:

Build the best travel guide

Thursday, 17 October 13

SLIDE 14

Mobile

Thursday, 17 October 13

SLIDE 15

Works offline

Thursday, 17 October 13

SLIDE 16

Covers the World

Thursday, 17 October 13

SLIDE 17

Smart

Thursday, 17 October 13

SLIDE 18

25,000 destinations

around the world

500,000

points of interest

~5,000,000

downloads

Thursday, 17 October 13

SLIDE 19

The Platform

Thursday, 17 October 13

SLIDE 20

Big data

Thursday, 17 October 13

SLIDE 21

Companies with Big data

Thursday, 17 October 13

SLIDE 22

Do you have big data?

Thursday, 17 October 13

SLIDE 23

We didn’t.

Thursday, 17 October 13

SLIDE 24

Nice, but pricey

Thursday, 17 October 13

SLIDE 25

Our current server room

Thursday, 17 October 13

SLIDE 26

Coders are more expensive than Python code is slow

Thursday, 17 October 13

SLIDE 27

Our System Architecture

Thursday, 17 October 13

SLIDE 28

Building A database of the world

Thursday, 17 October 13

SLIDE 29

Wikipedia OSM Wikitravel

Dropbox

Berlin

Uluru Potala Beansie

Google Spreadsheet

Buildguide Snapshot Pipeline Crawlers

The Flow

Thursday, 17 October 13

SLIDE 30

Crawl 20 Sources

Thursday, 17 October 13

SLIDE 31

Split everything in Pois & Locs

Thursday, 17 October 13

SLIDE 32

Put every thing back together

Thursday, 17 October 13

SLIDE 33

When are two things the same?

Thursday, 17 October 13

SLIDE 34

When they are: Similar in location and name

Thursday, 17 October 13

SLIDE 35

Wikpedia Wikitravel OSM

Geohashing

Thursday, 17 October 13

SLIDE 36

Suroit Camping vs Camping le Suroite

Thursday, 17 October 13

SLIDE 37

Shingling! Suroit Camping vs Camping le Suroite

Thursday, 17 October 13

SLIDE 38

ampi, camp, itca, mpin, oitc, ping, roit, suro, tcam, uroi

ampi, camp, esur, gles, ingl, lesu, mpin, ngle, oite, ping, roit, suro, uroi

60% Overlap

Thursday, 17 October 13

SLIDE 39

John’s Bar and Grill vs Restaurant John

Thursday, 17 October 13

SLIDE 40

Stop Shingles!

Thursday, 17 October 13

SLIDE 41

Haarlem Library vs Haarlem City Hall

Thursday, 17 October 13

SLIDE 42

Location Stop Shingles

Thursday, 17 October 13

SLIDE 43

Van Gogh Museum vs Van Gogh Hotel

Thursday, 17 October 13

SLIDE 44

Types from names

Thursday, 17 October 13

SLIDE 45

Cafe Sydney Opera House vs Sydney Opera House

Thursday, 17 October 13

SLIDE 46

Hmm

Thursday, 17 October 13

SLIDE 47

Side effect: Cuisine guesser

Thursday, 17 October 13

SLIDE 48

Making data useful

Thursday, 17 October 13

SLIDE 49

Some further processing...

Thursday, 17 October 13

SLIDE 50

Learning from pictures

Thursday, 17 October 13

SLIDE 51

Learning from Wikipedia

Thursday, 17 October 13

SLIDE 52

Wikipedia article distribution

Thursday, 17 October 13

SLIDE 53

Learning from ratings?

Thursday, 17 October 13

SLIDE 54

No Soup for you

Thursday, 17 October 13

SLIDE 55

Opinion Mining

Thursday, 17 October 13

SLIDE 56

Natural Language Processing

Thursday, 17 October 13

SLIDE 57

If you count RegExps...

Thursday, 17 October 13

SLIDE 58

http://labs.triposo.com

Further experiments

Thursday, 17 October 13

SLIDE 59

Can I suggest something?

Thursday, 17 October 13

SLIDE 60

The Weather

Thursday, 17 October 13

SLIDE 61

Your location

Text

Thursday, 17 October 13

SLIDE 62

Time

Thursday, 17 October 13

SLIDE 63

Personalization

Thursday, 17 October 13

SLIDE 64

Personalization

Thursday, 17 October 13

SLIDE 65

Personalization

Thursday, 17 October 13

SLIDE 66

Conclusion

Thursday, 17 October 13

SLIDE 67

Conclusion If you can’t solve a problem in 50 lines of python running on a server in the kitchen, it must be really hard.

Thursday, 17 October 13

SLIDE 68

Thanks!

Douwe Osinga (@dosinga)

Thursday, 17 October 13