Building the Inform Semantic Publishing Ecosystem: from Author to Audience
Download as PPT, PDF2 likes1,418 views
This document summarizes Inform, a content enrichment solution that uses semantic technologies to increase engagement for publishers. It analyzes content to extract topics, entities and related content which it then links and distributes through various channels. The summary focuses on how Inform processes content from authors, analyzes it semantically, and distributes related content to audiences through its platform and various distribution channels like widgets, microsites and social networks like Facebook.
Building the Inform Semantic Publishing Ecosystem: from Author to Audience
- 1. 1 More Meaning. Better Results. 1 Building the Inform Semantic Publishing Ecosystem: from Author to Audience Marc Hadfield VP, Research & Development marc@inform.com
- 2. 2 Marc Hadfield • Semantic Technology, Computer Science • Inform Technologies (Head of R&D) ‣ Semantic Technologies applied to Content Analysis & Distribution • Alitora Systems (Co-Founder / CTO) ‣ Life Science Semantic Technology, Research, Big Data Analytics, Semantic HPC ‣ Life Science Natural Language Processing • Columbia Genome Center ‣ NLP applied to Life Science Research Articles • LCconnect (CTO) ‣ Letter-of-Credit Exchange 2
- 3. 3 Semantics in Publishing… 3 • Ongoing Theme at ISWC 2010… ‣ NY Times ‣ Facebook (OpenGraph) ‣ Elsevier ‣ BBC
- 4. 4 What is Inform? 4 • Inform is a content enrichment solution designed to increase consumer engagement, page views and revenue. • We provide a hosted Semantic Web Service for content publishers that: 1. Reads your article before you publish it 2. Turns main topics and entities (people, places, companies, organizations) into links 3. Provides feeds of related web content when you publish it • New Direction: Optimizing Content Distribution via Direct Channels • Web users moving away from destination web sites, but still want the destination web site content. • Companies utilizing Inform include:
- 5. Connecting your content 55 Audio, Video & Blogs from the Web Articles from the Web Content from Inform Your Affiliates’ Content Your Content Affiliated Content Your Content Licensed Content Google Street View Topic 0.90 Google Company 1.00 Ireland Place 0.70 Norway Place 0.70 South Africa Place 0.70 Sweden Place 0.70 Brian McClendon Person 0.80 Mountain View, California Place 0.60 Wi-Fi Topic 0.50
- 7. 7 Inform Topic Pages, Micro Sites 7
- 8. 8 My Job: Building the Semantic Platform… 8 • “Silo”-ed Semantic Technology Semantic Web ‣ Aligned with Wikipedia, Leverage Linked Data for Mash-Ups ‣ RDFa, SKOS, Semantic SEO • Semantic / NLP Engine ‣ Improve Features, Quality • Semantic Data Infrastructure ‣ Scalable Infrastructure • Semantic Data Analysis ‣ Algorithms (Topology of Graphs), Inference ‣ “PageRank” on semantic data • Personalization, Usage Analysis • Micro Sites ‣ Clusters of Topics, Generating Rich Content Experience • Distributing to Social Platforms ‣ i.e. Facebook
- 9. 9 Inform: Author to Audience 9
- 11. 1111 Author ‣ Content Creation Services ‣ Semantic Data Repository ‣ Semantic Data Analysis ‣ Content Selection Algorithms ‣ Webservices ‣ Content Distribution Services Audience Inside the Semantic System Architecture
- 12. 12 Content Creation 12 • Article Creation Tool (ACT) ‣ Author Tools ‣ Embed in CMS, Tumblr / Wordpress Plugin • Publisher Portal ‣ Editorial Tool ‣ Content Feeds • Web Crawl • Summarizer ‣ Create smart “blurbs” to advertise article • LinkedData ‣ Freebase, Wikipedia, DBPedia, et cetera.
- 13. 13 ACT Tool 13
- 14. 14 ACT Tool 14
- 15. 15 ACT Tool, Tumblr, Wordpress 15
- 17. 17 Summarizer 17
- 18. 18 Semantic Data Repository 18 • Data Master / Data Node ‣ Federated Semantic Data Managers ‣ SPARQL Triplestore (scalable cluster) ‣ Semantic Search ‣ Search Indexes (Semi-Structured and Full-Text Search) ‣ Lucene/Siren (Sindice) ‣ Facets, Frequency Counts ‣ Cache (In-Memory) ‣ Blob Store (Voldemort) ‣ Listener to Activity (Flume) ‣ User Activity (clicks) ‣ Content Activity (content updates) ‣ Near Real-Time Trends, Analysis ‣ Compute Algorithms (Stored Procedures in Groovy) ‣ Long Term Content Archive (offline)
- 19. 19 Semantic Data Analysis 19 • Natural Language Processing ‣ Rules & Machine Learning, Training ‣ 500K articles per day, 4,000 unique sites ‣ Text Extraction, Section/Sentence Extraction ‣ Tokenization, Part-of-Speech, Noun/Verb Phrases ‣ Entity Extraction, Entity Normalization ‣ Topic Extraction, Summarization, Clustering • User Activity ‣ User Model (Personalization) • Semantic Inference ‣ F-Logic, Multi-Domain ‣ Linked Data Mash-Ups • Semantic Graph Topology ‣ Entity / Property Importance Metrics, Ranking, “PageRank” ‣ Which triples in LinkedData are interesting?
- 20. 20 Content Selection Algorithms 20 • Model of User, Personalization ‣ Social Networks provide Context • Semantic Analysis of Content • Algorithms ‣ Maximize Relevancy / Relatedness (Meets Editorial Criteria) ‣ Maximize Click-Through ‣ Cute Kitten vs. Engagement Issue ‣ Maximize Monetization Goal: Content Exchange
- 21. 21 Webservices 21 • REST ‣ Outputs RDF / JSON Data • Natural Language Processing ‣ Article to Semantic MetaData • Related Content ‣ Inputs: Content, Personalization, Algorithm ‣ Articles ‣ Semantic Mash-Ups ‣ Topics ‣ Entities • Semantic Query, Site Search • Storage, Content Repository
- 22. 22 Content Distribution Services 22 • Customer Destinations (Traditional Business) ‣ Deep Integration • Publisher Widgets ‣ Levels of Lightweight Integration ‣ Example: Related-Content-Widget in JavaScript • Inform.com ‣ Topic Pages • Micro Sites ‣ Several Thousand Owned-and-Operated Domains/Sites, Topic Driven • Social Networks ‣ Facebook Tools: • Semantic SEO ‣ RDFa, SKOS
- 25. 25 Using Facebook OpenGraph 25 Relevancy Algorithm: Combine: •Trending / Popular Topics •Trending / Popular Articles •Personalization “Liked” Topics •Personalization “Liked” Articles •User Profiles (“Users like you…”)
- 28. 28 Inform: Author to Audience via Semantics 28
- 29. 29 Thanks for your attention! 29 Questions? Contact Information: Marc Hadfield marc@inform.com