SlideShare a Scribd company logo
1 of 20
Application of NLG
in eCommerce
By Fatemeh Kazemi
Presented at TensorFlow User Group Toronto
Women in AI – 20 Mar 2019
1
Agenda
• What is NLG
• Types of NLG
• Template-Based NLG system
• Advance NLG system
• Sequence to Sequence model
• Encoder-Decoder architecture
• Application of Seq2Seq Model
• NLG in eCommerce
• ginnie.ai
• Model Architecture
• Results
• Challenges
• Solutions 2
What is NLG?
• NLG is characterized as the subfield of artificial
intelligence and computational linguistics that is
concerned with the construction of computer systems
than can produce understandable texts in English or other
human languages
• Template-based NLG
• Advanced NLG
3
NLG Type - Template-Based
• It uses templates with canned text and placeholders to
insert data into them
For example the systems that generate form letters
stating that a credit card spending limit is reached
• Advantages
Allow for full control over the quality of the output
Avoid the generation of ungrammatical structures
• Disadvantages:
Rely on hard-coded rules
Can not be easily reused across different projects
4
NLG Type - Advanced
• Advanced NLG systems automatically understand user
intent and generate content. They are intelligent
• Being very flexible thanks to the use of supervised and
unsupervised Machine Learning algorithms
• Using neural networks that learn morphological, lexical,
and grammar patterns from large corpora of written
language
• Allowing correcting language errors, such as misspellings
5
Seq2Seq Model
• The State-of-the-art NLG systems are built on deep neural
sequence to sequence models with an encoder-decoder
architecture
• Seq2seq models provide a framework to process
information that is in the form of sequences and both the
input and output are in the form of sequences of single
units like sequence of words, images, or speech units
• Convert sequences from one domain to sequences in
another domain
6
Encoder-Decoder Architecture
• Both the Encoder and Decoder are practically
two different neural network models
• The encoder is an RNN that reads each
symbol of an input sequence x sequentially
and as it reads each symbol, the hidden state
of the RNN changes
• After reading the end of the sequence
completely the hidden state of the RNN is a
summary of the whole input sequence
• Decoder is an RNN which is trained to
generate the output sequence by using the
hidden state of encoder 7
Applications of Seq2Seq Model
• Machine translation such as Google Translation
• Question answering application such as Siri
• Text summarization
• Speech Recognition
• Image captioning
• Video captioning
8
What Does NLG Mean For eCommerce?
• In a competitive environment where brands compete for wallet
share, the ability to create and present consistent content can be
a make-or-break
• A good description increases the likelihood of purchase
• It also helps your product gain visibility
• Usually product descriptions are written by writers or using
Template-based NLG tools
• By using Advanced NLG tools, businesses are able to generate
thousands of unique product descriptions in a matter of minutes
which are more cost effective than manual writing
• The advanced methods automatically transform raw product
specifications into engaging, unique, optimized content 9
Limited Content
10
ginnie.ai
11
ginnie.ai Example
NLG is characterized as ‘the subfield of artificial
intelligence and Inputs:
• Sleeve type: Long-Sleeve
• Pattern: Polka Dot
• Collar type: Classic Shirt
• Color: Black
• Material: Cotton
Output:
Complete your look with style this season in one of
these fun and fashionable tops for women. Opt for a
flattering and neutral black shade that is easy to
accessorize. A polka dot pattern adds whimsy and style
to this garment for a fun choice you'll enjoy wearing.
With a crisp, clean look, this classic shirt collar is a
staple in any woman's wardrobe. Keep arms protected
from the sun with these practical and comfortable long
sleeves. Cotton provides lasting comfort and retains its 12
Enhanced Content
13
Enhanced Content
14
ginnie Architecture Abstraction
• LSTM Encoder-Decoder is used for automatic description
generation
• LSTMs are a special kind of RNN, capable of learning long-term
dependencies
• We need 2 multilayered LSTM, one as encoder and another as
decoder
• The encoder processes the input sequence and returns its own
internal state
• The decoder is trained to turn the target sequences into the same
sequences but offset by one timestep in the future, this training
process called "teacher forcing"
• The decoder uses as initial state the state vectors from the
encoder, which is how the decoder obtains information about 15
Using Seq2Seq Model in ginnie
LSTM
Encoder
LSTM
Decoder
“Apparel Women Shirts Color Black”
“[START] Breathable and durable cotton materials allow for lightweight all-day comfort”
“Available in a dark, cool black for lots of attitude[STOP]”
“Apparel Women Shirts Material Cotton”
“Apparel Women Shirts Collar Classic”
“Apparel Women Shirts Sleeve Long-Sleeve” “[START] Get ready for the office in the fashionable long sleeves of this design”
“[START] Available in a dark, cool black for lots of attitude”
“[START] Designed with a classic shirt collar for a timeless look that is always in style”
“Breathable and durable cotton materials allow for lightweight all-day comfort[STOP]”
“Designed with a classic shirt collar for a timeless look that is always in style[STOP]”
“Get ready for the office in the fashionable long sleeves of this design[STOP]”
Internal LSTM
states(h,c)
16
ginnie Results
• We have obtained 80 to 90 percent accuracy across various
product categories
• We have achieved a reduction of 85 percent in time and
cost, resulting in us being able to produce high quality and
budget-friendly solutions for our customers
17
Challenges
• Lack of training data
Solutions
• Translating the sentences from one language to another
language
• Using Synonyms
18
References
1. C. Xiong, V. Zhong, and R. Socher, “Dynamic coattention networks for question answering,” in ICLR, 2016.
2. Cho, Kyunghyun, van Merrienboer, Bart, Gulcehre, Caglar, Bougares, Fethi, Schwenk, Holger, and Bengio, Yoshua.
Learning phrase representations using RNN encoder-decoder for statistical machine translation. In EMNLP, October
2014.
3. D. Bahdanau, J. Chorowski, D. Serdyuk, P. Brakel, and Y. Bengio, “End-to-end attention-based large vocabulary
speech recognition,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE,
2016, pp. 4945–4949.
4. Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. 2014. Sequence to sequence learning with neural networks. In NIPS.
5. J. Ba, V. Mnih, and K. Kavukcuoglu, “Multiple object recognition with visual attention,” arXiv preprint
arXiv:1412.7755, 2014.
6. K. Xu, J. Ba, R. Kiros, K. Cho, A. Courville, R. Salakhudinov, R. Zemel, and Y. Bengio, “Show, attend and tell: Neural
image caption generation with visual attention,” in International Conference on Machine Learning, 2015, pp. 2048–
2057.
7. Kyunghyun Cho, Bart van Merrienboer, C¸ aglar ¨ Gulc¸ehre, Dzmitry Bahdanau, Fethi Bougares, Hol- ¨ ger Schwenk,
and Yoshua Bengio. 2014. Learning phrase representations using rnn encoder-decoder for statistical machine
translation. In EMNLP.
8. R. Nallapati, B. Zhou, C. dos Santos, C. Gulcehre, and B. Xiang, “Abstractive text summarization using sequence-to-
sequence rnns and beyond,” in Proceedings of The 20th SIGNLL Conference on Computational Natural Language
Learning, 2016, pp. 280–290.
9. T. Luong, H. Pham, and C. D. Manning, “Effective approaches to attention-based neural machine translation,” in
Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2015, pp. 1412–1421. 19
Application of NLG in e commerce

More Related Content

What's hot

Easy and efficient processing of massive data sets
Easy and efficient processing of massive data setsEasy and efficient processing of massive data sets
Easy and efficient processing of massive data sets
jfrodrigues
 

What's hot (20)

Interpretability of Convolutional Neural Networks - Eva Mohedano - UPC Barcel...
Interpretability of Convolutional Neural Networks - Eva Mohedano - UPC Barcel...Interpretability of Convolutional Neural Networks - Eva Mohedano - UPC Barcel...
Interpretability of Convolutional Neural Networks - Eva Mohedano - UPC Barcel...
 
Chip packaging technology
Chip packaging technologyChip packaging technology
Chip packaging technology
 
Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)Convolutional Neural Networks (CNN)
Convolutional Neural Networks (CNN)
 
Intro To Convolutional Neural Networks
Intro To Convolutional Neural NetworksIntro To Convolutional Neural Networks
Intro To Convolutional Neural Networks
 
How to find defects in SMT electronics manufacturing
How to find defects in SMT electronics manufacturingHow to find defects in SMT electronics manufacturing
How to find defects in SMT electronics manufacturing
 
ALD/CVD applications, equipment and precursors in high volume manufacturing
ALD/CVD applications, equipment and precursors in high volume manufacturingALD/CVD applications, equipment and precursors in high volume manufacturing
ALD/CVD applications, equipment and precursors in high volume manufacturing
 
Transformers
TransformersTransformers
Transformers
 
Status of the Advanced Packaging Industry 2018 Report by Yole Developpement
Status of the Advanced Packaging Industry 2018 Report by Yole Developpement	Status of the Advanced Packaging Industry 2018 Report by Yole Developpement
Status of the Advanced Packaging Industry 2018 Report by Yole Developpement
 
How does your Stack-up, stack up?
How does your Stack-up, stack up?How does your Stack-up, stack up?
How does your Stack-up, stack up?
 
Attention is all you need (UPC Reading Group 2018, by Santi Pascual)
Attention is all you need (UPC Reading Group 2018, by Santi Pascual)Attention is all you need (UPC Reading Group 2018, by Santi Pascual)
Attention is all you need (UPC Reading Group 2018, by Santi Pascual)
 
R-CNN
R-CNNR-CNN
R-CNN
 
Going Global: How VICELAND Accelerated Channel Launches Around the World with...
Going Global: How VICELAND Accelerated Channel Launches Around the World with...Going Global: How VICELAND Accelerated Channel Launches Around the World with...
Going Global: How VICELAND Accelerated Channel Launches Around the World with...
 
Introduction to Capsule Networks (CapsNets)
Introduction to Capsule Networks (CapsNets)Introduction to Capsule Networks (CapsNets)
Introduction to Capsule Networks (CapsNets)
 
Relational knowledge distillation
Relational knowledge distillationRelational knowledge distillation
Relational knowledge distillation
 
Lip reading Project
Lip reading ProjectLip reading Project
Lip reading Project
 
Mtech Resume
Mtech ResumeMtech Resume
Mtech Resume
 
Introduction to Deep face detection and recognition
Introduction to Deep face detection and recognitionIntroduction to Deep face detection and recognition
Introduction to Deep face detection and recognition
 
Easy and efficient processing of massive data sets
Easy and efficient processing of massive data setsEasy and efficient processing of massive data sets
Easy and efficient processing of massive data sets
 
Transforming deep into transformers – a computer vision approach
Transforming deep into transformers – a computer vision approachTransforming deep into transformers – a computer vision approach
Transforming deep into transformers – a computer vision approach
 
Scorm Implementation Strategies for Mobile
Scorm Implementation Strategies for MobileScorm Implementation Strategies for Mobile
Scorm Implementation Strategies for Mobile
 

Similar to Application of NLG in e commerce

NetWork - 15.10.2011 - Applied code generation in .NET
NetWork - 15.10.2011 - Applied code generation in .NET NetWork - 15.10.2011 - Applied code generation in .NET
NetWork - 15.10.2011 - Applied code generation in .NET
Dmytro Mindra
 
Reading System for the Blind PPT
Reading System for the Blind PPTReading System for the Blind PPT
Reading System for the Blind PPT
Binayak Ghosh
 
Solving the System-Level Design Riddle
Solving the System-Level Design RiddleSolving the System-Level Design Riddle
Solving the System-Level Design Riddle
Design World
 

Similar to Application of NLG in e commerce (20)

OWF14 - Big Data : The State of Machine Learning in 2014
OWF14 - Big Data : The State of Machine  Learning in 2014OWF14 - Big Data : The State of Machine  Learning in 2014
OWF14 - Big Data : The State of Machine Learning in 2014
 
NetWork - 15.10.2011 - Applied code generation in .NET
NetWork - 15.10.2011 - Applied code generation in .NET NetWork - 15.10.2011 - Applied code generation in .NET
NetWork - 15.10.2011 - Applied code generation in .NET
 
NLP and Deep Learning for non_experts
NLP and Deep Learning for non_expertsNLP and Deep Learning for non_experts
NLP and Deep Learning for non_experts
 
Digital Library Solutions
Digital Library SolutionsDigital Library Solutions
Digital Library Solutions
 
Precision Content™ Tools, Techniques, and Technology
Precision Content™ Tools, Techniques, and TechnologyPrecision Content™ Tools, Techniques, and Technology
Precision Content™ Tools, Techniques, and Technology
 
Reading System for the Blind PPT
Reading System for the Blind PPTReading System for the Blind PPT
Reading System for the Blind PPT
 
Age of Language Models in NLP
Age of Language Models in NLPAge of Language Models in NLP
Age of Language Models in NLP
 
AI hype or reality
AI  hype or realityAI  hype or reality
AI hype or reality
 
Understanding deep learning
Understanding deep learningUnderstanding deep learning
Understanding deep learning
 
Hello World - Introduction to coding.pptx
Hello World - Introduction to coding.pptxHello World - Introduction to coding.pptx
Hello World - Introduction to coding.pptx
 
Artificial Intelligence in practice - Gerbert Kaandorp - Codemotion Amsterdam...
Artificial Intelligence in practice - Gerbert Kaandorp - Codemotion Amsterdam...Artificial Intelligence in practice - Gerbert Kaandorp - Codemotion Amsterdam...
Artificial Intelligence in practice - Gerbert Kaandorp - Codemotion Amsterdam...
 
Livescribe Skywifi smartpen
Livescribe Skywifi smartpenLivescribe Skywifi smartpen
Livescribe Skywifi smartpen
 
An Introduction to Natural Language Processing
An Introduction to Natural Language ProcessingAn Introduction to Natural Language Processing
An Introduction to Natural Language Processing
 
Advanced Natural Language Processing with Apache Spark NLP
Advanced Natural Language Processing with Apache Spark NLPAdvanced Natural Language Processing with Apache Spark NLP
Advanced Natural Language Processing with Apache Spark NLP
 
5 pen technology
5 pen technology5 pen technology
5 pen technology
 
AI & Machine Learning Pipelines with Knative
AI & Machine Learning Pipelines with KnativeAI & Machine Learning Pipelines with Knative
AI & Machine Learning Pipelines with Knative
 
Cerebro for creative teams
Cerebro for creative teamsCerebro for creative teams
Cerebro for creative teams
 
Choosing the right Technologies for your next unicorn.
Choosing the right Technologies for your next unicorn.Choosing the right Technologies for your next unicorn.
Choosing the right Technologies for your next unicorn.
 
2014 12-10 aimee
2014 12-10 aimee2014 12-10 aimee
2014 12-10 aimee
 
Solving the System-Level Design Riddle
Solving the System-Level Design RiddleSolving the System-Level Design Riddle
Solving the System-Level Design Riddle
 

More from Devatanu Banerjee

More from Devatanu Banerjee (7)

TensorFlow 2.0 Autographs - For TFUG - Vik Pant
TensorFlow 2.0 Autographs - For TFUG - Vik PantTensorFlow 2.0 Autographs - For TFUG - Vik Pant
TensorFlow 2.0 Autographs - For TFUG - Vik Pant
 
TensorFlow 2.0 Overview - For TensorFlow User Group - Ehsan Amjadian
TensorFlow 2.0 Overview - For TensorFlow User Group - Ehsan AmjadianTensorFlow 2.0 Overview - For TensorFlow User Group - Ehsan Amjadian
TensorFlow 2.0 Overview - For TensorFlow User Group - Ehsan Amjadian
 
Font classification with 5 deep learning models using tensor flow
Font classification with 5 deep learning models using tensor flowFont classification with 5 deep learning models using tensor flow
Font classification with 5 deep learning models using tensor flow
 
Dave D'Silva - Emerging Technology Framework to Unite Society
Dave D'Silva - Emerging Technology Framework to Unite SocietyDave D'Silva - Emerging Technology Framework to Unite Society
Dave D'Silva - Emerging Technology Framework to Unite Society
 
Sandeep Chaudhury - AI & Prosperity
Sandeep Chaudhury - AI & ProsperitySandeep Chaudhury - AI & Prosperity
Sandeep Chaudhury - AI & Prosperity
 
Tensorflow User Group Toronto - Ehsan Amjadian - TF Gager
Tensorflow User Group Toronto - Ehsan Amjadian - TF GagerTensorflow User Group Toronto - Ehsan Amjadian - TF Gager
Tensorflow User Group Toronto - Ehsan Amjadian - TF Gager
 
Monitoring Trends (Seminar Slides)
Monitoring Trends (Seminar Slides)Monitoring Trends (Seminar Slides)
Monitoring Trends (Seminar Slides)
 

Recently uploaded

Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotecAbortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
Abortion pills in Riyadh +966572737505 get cytotec
 
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
aqpto5bt
 
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
acoha1
 
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
Amil baba
 
Displacement, Velocity, Acceleration, and Second Derivatives
Displacement, Velocity, Acceleration, and Second DerivativesDisplacement, Velocity, Acceleration, and Second Derivatives
Displacement, Velocity, Acceleration, and Second Derivatives
23050636
 
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
yulianti213969
 
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
zifhagzkk
 
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptxAudience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Stephen266013
 
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
ppy8zfkfm
 
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
pwgnohujw
 

Recently uploaded (20)

Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotecAbortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
Abortion pills in Riyadh Saudi Arabia (+966572737505 buy cytotec
 
Credit Card Fraud Detection: Safeguarding Transactions in the Digital Age
Credit Card Fraud Detection: Safeguarding Transactions in the Digital AgeCredit Card Fraud Detection: Safeguarding Transactions in the Digital Age
Credit Card Fraud Detection: Safeguarding Transactions in the Digital Age
 
MATERI MANAJEMEN OF PENYAKIT TETANUS.ppt
MATERI  MANAJEMEN OF PENYAKIT TETANUS.pptMATERI  MANAJEMEN OF PENYAKIT TETANUS.ppt
MATERI MANAJEMEN OF PENYAKIT TETANUS.ppt
 
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
一比一原版(ucla文凭证书)加州大学洛杉矶分校毕业证学历认证官方成绩单
 
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
如何办理(WashU毕业证书)圣路易斯华盛顿大学毕业证成绩单本科硕士学位证留信学历认证
 
Aggregations - The Elasticsearch "GROUP BY"
Aggregations - The Elasticsearch "GROUP BY"Aggregations - The Elasticsearch "GROUP BY"
Aggregations - The Elasticsearch "GROUP BY"
 
The Significance of Transliteration Enhancing
The Significance of Transliteration EnhancingThe Significance of Transliteration Enhancing
The Significance of Transliteration Enhancing
 
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
NO1 Best Kala Jadu Expert Specialist In Germany Kala Jadu Expert Specialist I...
 
Displacement, Velocity, Acceleration, and Second Derivatives
Displacement, Velocity, Acceleration, and Second DerivativesDisplacement, Velocity, Acceleration, and Second Derivatives
Displacement, Velocity, Acceleration, and Second Derivatives
 
Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
Identify Customer Segments to Create Customer Offers for Each Segment - Appli...Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
Identify Customer Segments to Create Customer Offers for Each Segment - Appli...
 
Bios of leading Astrologers & Researchers
Bios of leading Astrologers & ResearchersBios of leading Astrologers & Researchers
Bios of leading Astrologers & Researchers
 
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
obat aborsi Bontang wa 081336238223 jual obat aborsi cytotec asli di Bontang6...
 
NOAM AAUG Adobe Summit 2024: Summit Slam Dunks
NOAM AAUG Adobe Summit 2024: Summit Slam DunksNOAM AAUG Adobe Summit 2024: Summit Slam Dunks
NOAM AAUG Adobe Summit 2024: Summit Slam Dunks
 
Predictive Precipitation: Advanced Rain Forecasting Techniques
Predictive Precipitation: Advanced Rain Forecasting TechniquesPredictive Precipitation: Advanced Rain Forecasting Techniques
Predictive Precipitation: Advanced Rain Forecasting Techniques
 
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
如何办理(Dalhousie毕业证书)达尔豪斯大学毕业证成绩单留信学历认证
 
Audience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptxAudience Researchndfhcvnfgvgbhujhgfv.pptx
Audience Researchndfhcvnfgvgbhujhgfv.pptx
 
How to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data AnalyticsHow to Transform Clinical Trial Management with Advanced Data Analytics
How to Transform Clinical Trial Management with Advanced Data Analytics
 
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
1:1原版定制利物浦大学毕业证(Liverpool毕业证)成绩单学位证书留信学历认证
 
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarjSCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
SCI8-Q4-MOD11.pdfwrwujrrjfaajerjrajrrarj
 
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
原件一样(UWO毕业证书)西安大略大学毕业证成绩单留信学历认证
 

Application of NLG in e commerce

  • 1. Application of NLG in eCommerce By Fatemeh Kazemi Presented at TensorFlow User Group Toronto Women in AI – 20 Mar 2019 1
  • 2. Agenda • What is NLG • Types of NLG • Template-Based NLG system • Advance NLG system • Sequence to Sequence model • Encoder-Decoder architecture • Application of Seq2Seq Model • NLG in eCommerce • ginnie.ai • Model Architecture • Results • Challenges • Solutions 2
  • 3. What is NLG? • NLG is characterized as the subfield of artificial intelligence and computational linguistics that is concerned with the construction of computer systems than can produce understandable texts in English or other human languages • Template-based NLG • Advanced NLG 3
  • 4. NLG Type - Template-Based • It uses templates with canned text and placeholders to insert data into them For example the systems that generate form letters stating that a credit card spending limit is reached • Advantages Allow for full control over the quality of the output Avoid the generation of ungrammatical structures • Disadvantages: Rely on hard-coded rules Can not be easily reused across different projects 4
  • 5. NLG Type - Advanced • Advanced NLG systems automatically understand user intent and generate content. They are intelligent • Being very flexible thanks to the use of supervised and unsupervised Machine Learning algorithms • Using neural networks that learn morphological, lexical, and grammar patterns from large corpora of written language • Allowing correcting language errors, such as misspellings 5
  • 6. Seq2Seq Model • The State-of-the-art NLG systems are built on deep neural sequence to sequence models with an encoder-decoder architecture • Seq2seq models provide a framework to process information that is in the form of sequences and both the input and output are in the form of sequences of single units like sequence of words, images, or speech units • Convert sequences from one domain to sequences in another domain 6
  • 7. Encoder-Decoder Architecture • Both the Encoder and Decoder are practically two different neural network models • The encoder is an RNN that reads each symbol of an input sequence x sequentially and as it reads each symbol, the hidden state of the RNN changes • After reading the end of the sequence completely the hidden state of the RNN is a summary of the whole input sequence • Decoder is an RNN which is trained to generate the output sequence by using the hidden state of encoder 7
  • 8. Applications of Seq2Seq Model • Machine translation such as Google Translation • Question answering application such as Siri • Text summarization • Speech Recognition • Image captioning • Video captioning 8
  • 9. What Does NLG Mean For eCommerce? • In a competitive environment where brands compete for wallet share, the ability to create and present consistent content can be a make-or-break • A good description increases the likelihood of purchase • It also helps your product gain visibility • Usually product descriptions are written by writers or using Template-based NLG tools • By using Advanced NLG tools, businesses are able to generate thousands of unique product descriptions in a matter of minutes which are more cost effective than manual writing • The advanced methods automatically transform raw product specifications into engaging, unique, optimized content 9
  • 12. ginnie.ai Example NLG is characterized as ‘the subfield of artificial intelligence and Inputs: • Sleeve type: Long-Sleeve • Pattern: Polka Dot • Collar type: Classic Shirt • Color: Black • Material: Cotton Output: Complete your look with style this season in one of these fun and fashionable tops for women. Opt for a flattering and neutral black shade that is easy to accessorize. A polka dot pattern adds whimsy and style to this garment for a fun choice you'll enjoy wearing. With a crisp, clean look, this classic shirt collar is a staple in any woman's wardrobe. Keep arms protected from the sun with these practical and comfortable long sleeves. Cotton provides lasting comfort and retains its 12
  • 15. ginnie Architecture Abstraction • LSTM Encoder-Decoder is used for automatic description generation • LSTMs are a special kind of RNN, capable of learning long-term dependencies • We need 2 multilayered LSTM, one as encoder and another as decoder • The encoder processes the input sequence and returns its own internal state • The decoder is trained to turn the target sequences into the same sequences but offset by one timestep in the future, this training process called "teacher forcing" • The decoder uses as initial state the state vectors from the encoder, which is how the decoder obtains information about 15
  • 16. Using Seq2Seq Model in ginnie LSTM Encoder LSTM Decoder “Apparel Women Shirts Color Black” “[START] Breathable and durable cotton materials allow for lightweight all-day comfort” “Available in a dark, cool black for lots of attitude[STOP]” “Apparel Women Shirts Material Cotton” “Apparel Women Shirts Collar Classic” “Apparel Women Shirts Sleeve Long-Sleeve” “[START] Get ready for the office in the fashionable long sleeves of this design” “[START] Available in a dark, cool black for lots of attitude” “[START] Designed with a classic shirt collar for a timeless look that is always in style” “Breathable and durable cotton materials allow for lightweight all-day comfort[STOP]” “Designed with a classic shirt collar for a timeless look that is always in style[STOP]” “Get ready for the office in the fashionable long sleeves of this design[STOP]” Internal LSTM states(h,c) 16
  • 17. ginnie Results • We have obtained 80 to 90 percent accuracy across various product categories • We have achieved a reduction of 85 percent in time and cost, resulting in us being able to produce high quality and budget-friendly solutions for our customers 17
  • 18. Challenges • Lack of training data Solutions • Translating the sentences from one language to another language • Using Synonyms 18
  • 19. References 1. C. Xiong, V. Zhong, and R. Socher, “Dynamic coattention networks for question answering,” in ICLR, 2016. 2. Cho, Kyunghyun, van Merrienboer, Bart, Gulcehre, Caglar, Bougares, Fethi, Schwenk, Holger, and Bengio, Yoshua. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In EMNLP, October 2014. 3. D. Bahdanau, J. Chorowski, D. Serdyuk, P. Brakel, and Y. Bengio, “End-to-end attention-based large vocabulary speech recognition,” in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 2016, pp. 4945–4949. 4. Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. 2014. Sequence to sequence learning with neural networks. In NIPS. 5. J. Ba, V. Mnih, and K. Kavukcuoglu, “Multiple object recognition with visual attention,” arXiv preprint arXiv:1412.7755, 2014. 6. K. Xu, J. Ba, R. Kiros, K. Cho, A. Courville, R. Salakhudinov, R. Zemel, and Y. Bengio, “Show, attend and tell: Neural image caption generation with visual attention,” in International Conference on Machine Learning, 2015, pp. 2048– 2057. 7. Kyunghyun Cho, Bart van Merrienboer, C¸ aglar ¨ Gulc¸ehre, Dzmitry Bahdanau, Fethi Bougares, Hol- ¨ ger Schwenk, and Yoshua Bengio. 2014. Learning phrase representations using rnn encoder-decoder for statistical machine translation. In EMNLP. 8. R. Nallapati, B. Zhou, C. dos Santos, C. Gulcehre, and B. Xiang, “Abstractive text summarization using sequence-to- sequence rnns and beyond,” in Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning, 2016, pp. 280–290. 9. T. Luong, H. Pham, and C. D. Manning, “Effective approaches to attention-based neural machine translation,” in Proceedings of the Conference on Empirical Methods in Natural Language Processing, 2015, pp. 1412–1421. 19