wechat - social to intelligent connection...text/voice answer speech recognition tts nlg data...
TRANSCRIPT
WeChat Social to Intelligent Connection
杨强
What is WeChat?
A New Lifestyle
• Mobile App for 500M+
• Multimedia
• A Way to Connect: A Platform
Rapid Growth
0
2
4
6
8
10
12
1.0 4.68
Oversea Users By Aug 2013
Active Users per Month
By Sep 2014
> 1 0
Users(Unit:100M)
1 Billion Accounts, 550M Active Users
8M Service Accts
20 Languages, 200 Countries
微信 Growing Up Path
Birth
Speech
Nearby
Shake It!
Scan
Service Platform
Video Chat
Payment Online Shopping
Red Packet
2014-10-1
Video Clips
Moment Ads
1 2
3
2012-3-29
2012-9-17
2013-1-15
Growth Path (Unit: 100M)
~6
2015
2010
2011
2012
2013
2014
微信诞生
-扫一扫
-视频聊天
-支付
-游戏
-表情商店
-红包
-小视频
-卡券
-广告
1亿
2亿
3亿
2015
-语音信息
-附近的人
-摇一摇
-漂流瓶
-朋友圈
-公众平台
2012-3-29
2012-9-17
2013-1-15
2014-10-1
433 days: from 0 to 100 M
6 Months Later: doubled to 200M
3 months later: 300 M
3 Milestones
COLLECTIVE USER EXPERIENCE: SHAKING FOR RED POCKET IN 2015 SPRING FESTIVAL
Red envelope data:Over 1.01 billion times、Peak at 550 000 / min、Shaking 11
billion times, Peak 810 million times / min
• INTERACTION / COMMENTS / VOTING / LOTTERY / SHARE WITH FRIENDS
LINK
BigData @ WeChat
WeChat Big Data
Big
Dat
a
relational data
contacts
subscribed public accounts
group
non-relational
explicit information
behavior
posts
text
images
videos comments user feedback
Relationship Types (%)
陌生人
网友
老师或领导
同事
亲人或亲戚
同学
现实生活中的朋友
24.7
32.1
50.4
70.8
75.7
81.4
90
Friends in Real Life
Strangers
Classmates
Artificial Intelligence @
Image Understanding Lifelong Learning Agent Spammer and Rumor
detection
Location-based social
networks
NLP and Text Recommendation Feature Engineering Provence of information
Speech Understanding Social Search Trust Assessment User Modeling
Crowd Intelligence Sentiment Analysis Event Detection Network Analysis
WeChat User Modeling & Transfer Learning
Case Study: ADs IN WeChat MOMENTS
• PRESENTS IN MOMENTS / TARGETING BASED ON BIG
DATA / INNTERACTIVE COMMUNICATION
• SPREADING BETWEEN FRIENDS
User Modeling for Ads
• Data Sources • Demographic information
• Articles read
• Public accounts subscribed
• Techniques: • DNN
• Multi-task Learning
Source 1
Source 2
Source 3
Source 1
Source 2
Source 3
Topic modelling
Topic modelling
Topic modelling
Co-training
tag 1
tag n
Seed users
Advertise
Image Understanding for Ads
Cross Domain Transfer Learning
• Predicting User Feedback from Social Data
source domain: BMD advertisement with abundant feedbacks
Target domain: SOHO advertising with no previous data
Source domain: BMW advertisement with user feedback
Crowd Intelligence @ WeChat
2010
2011
2012
2013
2014
微信诞生
-扫一扫
-视频聊天
1亿
2亿
3亿
2015
-语音信息
-附近的人
-摇一摇
-漂流瓶
-红包
-游戏
-表情商店
2012-3-29
2012-9-17
2013-1-15
-朋友圈
-公众平台
-小视频
-卡券
-广告
Grow with Users
- Games
- Red Pocket
- Moments Ads
- Shaking TV Program
Crowd Intelligence
Charity by the Millions: Voice Donor
LINK
Crowd Intelligence for the Visually Impaired
Collect Voices Filter by Standard Models
Speech Recognition
• Large Mandarin Corpus: DNN (deep neutral network)
• Language model: • N-gram, DNN • Low-rank matrix • GPU training
• Decoder: • WFST framework • Large, parallel search space
Audio Fingerprinting
• Challenge • noisy environments,
• compactness of fingerprint, and
• service scalability when song database is huge (10M)
• Application: WeChat “Shake” Music, lunched in Jan, 2013
• Big Music Database(10M songs) Fast Recognition (3-5 seconds)
• Daily Page View > 8M, User View > 3M
Audio Fingerprinting:WeChat Live TV recognition
Recognize live TV program from audio
fingerprinting
• Challenge: High concurrent throughput
• SHAKE-TV:
• Can recognize > 500 TV channels
across China
• User View: 1M simultaneously
• Rich Cross-TV Screen User Experience
• Fully integrated with social networks
Image Understanding @ WeChat
Configuration Speed-up
2 GPUs Model
Parallelism
1.71
2 GPUs Data Parallelism 1.85
4 GPUs Model + Data
Parallelism
2.52
4 GPUs Data Parallelism 2.67
GPU0
GPU1
GPU2
GPU3
Mariana CNN on GPUs
Mariana CNN is Tencent’s Deep Convolutional Neural Network based on Single-machine, Multi-core GPU Computation.
• Data parallelism and model parallelism
• Partition models for parallel execution
• Model scalability and performance had major improvement
• Scan for information or services
• Local and Global Image Feature Descriptor
• Highly Efficient Feature Indexing and Matching
• Mobile video and image quality assessment
• Challenges:
• Variable lighting, Non-planar recognition
• WeChat “Scan” on covers, lunched in 2013
• Large Image databases (~10M)
• Open interfaces for developers
Mobile Visual Search
LINK
• OCR on a mobile device
• Camera OCR based language Translation
• Certificate and ID OCR on mobile or cloud
• Face Technology:
• Detection, Alignment, Tracking, Recognition/Verification
• User modeling based on images
• Targeted Adverts
Mobile Image Tech.
LINK
Augmented Reality
3D animation w/ embedded video on designs • Rich interaction with users
Challenges: • Real-time and precise target
detecting/tracking, model rendering
Applications: • lunched in WeChat movie ticket App “微票” (Jan 2015)
LINK
Natural Language Understanding @ WeChat
WeChat: Closed-loop NLP • Closed-loop Feedback in WeChat Services
• Always online: real-time message platform
• Massive user base: 549 million monthly active users
• Payment User Intention
…
payment
WeChat NLP
Word Multi-Embedding
Learn Embedded Word Representation
WeChat NLP
Semantic Matching of Questions and Answers
h
output
R(query,doc1)
WARP loss
query doc1 doc2
R(query,doc2)
h x
x h
x x
h
h x
x h
x x
h
h x
h x
x x
output output
Dependency-Tree RNN model
• semantic match
• semantic answer ranking
Semantic Match
• Dependency-Tree RNN
•Word multi-embedding match
•BM25
•Other Features:
-Sentence type recognition
-Synonym
-Antonym
-Parsing
WeChat NLP
Text/Voice
Answer
Speech Recognition
TTS
Data management
Pattern based
Parsing and LM based
NLU Query Analysis
Query Inference
Semantic Pattern
intent identification
Self-learning system
query rewrite NER/NED Parsing
Sentiment Analysis Graph Search
RDF based Inverted index based
NLG
Log/Session
management
knowledge
graph
Dialog
management
User Profile
User behavior
Context
management
Semantic Search
Semantic Match
ReRanking
Inverted index
Text Search
Recommendation
NLP & Question Answering
NLP Dialog
微信 Future
Connect Everything
A New Connection Model
Becomes a New Lifestyle
Will extend the connection to daily life, commerce &
entertainment
Provide Mobile Internet Service to Industries
People
Service Things
Connecting People, Services and Things
People, things and services An ecosystem for connection, a new solution provider
Internet +
Intelligent Business Solutions
Integrate internet and other businesses,Smart
Cities,Improved User experience, Connecting
everything
Eight Principles
1、Bring New Value
2、Remove geographical restrictions
3、Remove middle men
4、Distributed
5、Ecosystem
6、Evolutionary Service Platform
7、Social Centric
8、Users’ Interests Always #1
THANK YOU