beyond keywords june 2015
TRANSCRIPT
Beyond Keywords: The Revolution in Search
SLA Contributed Papers, Session 1885
June 15, 2015
Joe Buzzanga, Senior Product Manager, IEEE
Kurzweil's Law of Accelerating Returns
"An analysis of the history of technology shows that technological change is exponential, contrary to the common-sense “intuitive linear” view".http://www.kurzweilai.net/the-law-of-accelerating-returns
What is Search?
Search: Basic Model
The Dream of Search
Yet, in many ways, we’re a million miles away from creating the search engine of my dreams, one that gets you just the right information at the exact moment you need it with almost no effort.
Larry Page, 2013 Google Founders Letter, https://investor.google.com/corporate/2013/founders-letter.html
Recipe for a Dream
AI (especially machine learning) +
Knowledge Bases +
Intelligent Personal Assistants
Achieving the Dream
Search = AIAI-->Machine Learning
Machine Learning-->Deep Neural Networks
Machine Learning (ML)
Kulkarni, Parag. "Introduction to reinforcement and systemic machine learning."Reinforcement and Systemic Machine Learning for Decision Making (2012)
Machine Learning (ML) is EverywhereBaidu’s Artificial-Intelligence Supercomputer Beats Google at Image RecognitionMay 13, 2015 http://www.technologyreview.com/
Scientists See Promise in Deep-Learning ProgramsNov. 23, 2012 http://www.nytimes.com/2012/11/24/science/
A Google Computer Can Teach Itself GamesFeb. 25, 2015 http://bits.blogs.nytimes.com/
Amazon's Fire Phone recognizes everything around you with 'Firefly'By Chris Welch on June 18, 2014http://www.theverge.com
Facebook Launches Advanced AI Effort to Find Meaning in Your PostsSept. 2013 http://www.technologyreview.com
Google Acquires British Artificial Intelligence DeveloperJan. 27, 2014 http://dealbook.nytimes.com
Amazon Launches Machine Learning-As-A-ServiceApril 10, 2015 http://www.informationweek.com
ML is Everywhere
"Google using ML in 47 products" Jeff Dean, 2015 NVIDIA GPU Technology Conference
ML is Everywhere
ML (Neural Nets) for Text
This article demonstrates that we can apply deep learning to text understanding from character level inputs all the way up to abstract text concepts, using temporal convolutional networks(LeCun et al., 1998) (ConvNets). We apply ConvNets to various large-scale datasets, including ontology classification, sentiment analysis, and text categorization. We show that temporal ConvNets can achieve astonishing performance without the knowledge of words, phrases, sentences and any other syntactic or semantic structures with regards to a human language. Evidence shows that our models can work for both English and Chinese.
arXiv:1502.01710v2 [cs.LG] 7 Apr 2015
Text Understanding from Scratch
ML (Neural Nets) for Text
Natural Language Processing (Almost) from ScratchWe propose a unified neural network architecture and learning algorithm that can be applied to various natural language processing tasks including part-of-speech tagging, chunking, named entity recognition, and semantic role labeling…. our system learns internal representations on thebasis of vast amounts of mostly unlabeled training data.
Journal of Machine Learning Research 12 (2011) 2461-2505 Submitted 1/10; Revised 11/10; Published 8/11
Using ML in Search: Patent Search Example
In this invention, we propose to utilize analog behaviors of off-the-shelf Flash memory to enable hardware-based security functions in a wide range of electronic devices without requiring custom hardware. More specifically, we show that a standard Flash memory interface can be used to generate true random numbers from quantum and thermal noises and to produce device fingerprints based on manufacturing variations. The techniques can be applied to any floating-gate non-volatile memory in general, and does not require any hardware modifications to today's Flash memory chips, allowing them to be widely deployed.
ML--Topic Models
Knowledge Bases Emerge
Knowledge Base: Application
• Understand query
• Understand content (web pages, documents, images…)
• Return answers
• Inference?"We’ve gone up a level from just talking about the words to talking about what the thing actually is. In crawling and indexing documents we can now have an understanding of what the document is about. If the document is about famous tennis players we actually know it’s about sport and tennis. Every piece of information that we crawl, index, or search is analyzed in the context of Knowledge Graph. That’s not the same as completely understanding the text as you and I might do but it’s a step towards it.--John Giannandrea, Google VP, Jan. 2014 http://www.technologyreview.com
Knowledge Base Comparison
Name Entities Facts Source
DBpedia 4.6M 583M Wikipedia
YAGO 10M 120M Wikipedia, WordNet, GeoNames
NELL 5.2M 50M Candidates, 2.4M Confident
Machine learning, crawls open web
Knowledge Graph (Google)
500M 3.5B Wikipedia, Freebase, and other sources
Knowledge Vault (Google)
45M 271M Machine learning, crawls open web
Knowledge Base: Examples
Knowledge Base: Examples
Intelligent Personal Assistants
Intelligent Personal Assistants
How well do they perform?One recent report tested the question answering ability of Google Now, Siri and Cortana, based on over 3,000 queries. Google Now returned a direct, correct answer to 51% of the queries, while Siri managed to answer 15% correctly and Cortana only 8%.
Vendor Product Initial Release Scope
Apple Siri October, 2011 General
Google Google Now July, 2012 General
Microsoft Cortana April, 2014 General
Samsung S Voice May, 2012 General
Nuance Ask Nina August, 2012Task specific – Customer care
x.ai x.ai Restricted betaTask specific – Scheduling assistant
Viv Viv In development General
Conclusions
• User expectations are rising
• Seize opportunities to add value beyond the traditional
• Seek out Data Science expertise
• Partner with IT
• Look for ML-savvy vendors