Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
A Natural Language Pipeline
Search
ddqz
July 06, 2019
Technology
0
500
A Natural Language Pipeline
Presentation from the spaCy IRL 2019 conference.
ddqz
July 06, 2019
Tweet
Share
Other Decks in Technology
See All in Technology
United Airlines Customer Service– Call 1-833-341-3142 Now!
airhelp
0
170
【5分でわかる】セーフィー エンジニア向け会社紹介
safie_recruit
0
27k
「クラウドコスト絶対削減」を支える技術—FinOpsを超えた徹底的なクラウドコスト削減の実践論
delta_tech
4
170
Model Mondays S2E04: AI Developer Experiences
nitya
0
140
オーティファイ会社紹介資料 / Autify Company Deck
autifyhq
10
130k
生成AI活用の組織格差を解消する 〜ビジネス職のCursor導入が開発効率に与えた好循環〜 / Closing the Organizational Gap in AI Adoption
upamune
7
5.3k
品質と速度の両立:生成AI時代の品質保証アプローチ
odasho
1
360
Delegating the chores of authenticating users to Keycloak
ahus1
0
140
Flutter向けPDFビューア、pdfrxのpdfium WASM対応について
espresso3389
0
130
怖くない!はじめてのClaude Code
shinya337
0
400
DatabricksにOLTPデータベース『Lakebase』がやってきた!
inoutk
0
110
IPA&AWSダブル全冠が明かす、人生を変えた勉強法のすべて
iwamot
PRO
2
140
Featured
See All Featured
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
46
9.6k
The Power of CSS Pseudo Elements
geoffreycrofte
77
5.9k
Balancing Empowerment & Direction
lara
1
430
It's Worth the Effort
3n
185
28k
Docker and Python
trallard
44
3.5k
Testing 201, or: Great Expectations
jmmastey
43
7.6k
Making Projects Easy
brettharned
116
6.3k
Done Done
chrislema
184
16k
GitHub's CSS Performance
jonrohan
1031
460k
ピンチをチャンスに:未来をつくるプロダクトロードマップ #pmconf2020
aki_iinuma
126
53k
Into the Great Unknown - MozCon
thekraken
40
1.9k
We Have a Design System, Now What?
morganepeng
53
7.7k
Transcript
A Natural Language Pipeline
More Input
Knowledge” “A compendium of human...
Library
Physical archives became digital records, encoded with metadata
The internet promised rich dynamic experiences
The internet promised rich dynamic experiences but served us banner
ads
Advertising has and continues to fuel a substantial portion of
the innovation on the internet
What would The Economist look like if it were founded
in 2012?
User
First
Experience
“There’s a reason that tech companies are topping the lists
of most valuable companies and brands. Every company is a tech company.” Maggie Chan Jones
Every story, at its core, is a business story
Language
None
None
Stage -> Stenographer -> Editors -> spaCy -> Data Store
<-> Backend <- Slack <- Users Proto-Pipeline
Over eight hours we created data from the content of
the event, building the model in real-time
The model evolved over time
This was the experiment that would evolve into SiO 2
Silicon, a key element in everything from glass to microchips,
is at the core of global business
Oxygen, the journalistic voice Quartz breathes into the global business
news cycle
Entities are linguistic anchors, defined by context and around which
context can be inferred
Standard Entities PERSON FACILITY ORG PRODUCT GPE EVENT... Additional Entities
TECHNOLOGY PROCESS NATURE MEDIA CONSTRUCT
70K articles 1.4M blocks of text 85K labeled sentences
Entities
This spaCy model made rich analysis for any given text
easy to do on the fly
Stored analysis of a large corpus is a vital resource
The language graph...
Graph
The language graph is a mutable map of the language
model
Any new content is analyzed and then mapped onto the
language graph
Changes made to the graph can then be incorporated into
the next model iteration
The language graph becomes a primary resource for extracting training
data
Snapshots of time can be extracted from the language graph
Context can be derived by looking at the relationships in
the language graph
Elon Musk
Jeff Bezos
Mark Zuckerberg
Context
SiO 2 is a living Natural Language Pipeline of networked
algorithms trained on the corpus of Quartz to understand the linguistic patterns of global business news
The Pipeline(s) Quartz Corpus -> Training Sentences -> spaCy Content
-> spaCy -> Language Graph Language Graph -> Training Data -> Statistical Models / Classifiers Language Graph -> Training Sentences -> spaCy Unseen Content -> spaCy -> Pre-Processed Text / Vectors -> Statistical Models / Classifiers
Thank you