Semantic Analytics at Scale

Andrew Montalenti (co-founder & CTO of Parse.ly) presents "Semantic Analytics at Scale" at NYTimes Headquarters. For the TimesOpen event, "Bigger Data and Smarter Scaling".

Related links:

Parse.ly: http://parse.ly
Schema.to: http://schema.to
Live blog of event: http://open.blogs.nytimes.com/2012/10/17/live-blogging-timesopen-bigger-data-and-smarter-scaling/

Transcript

1. Semantic Analytics at Scale
   Wednesday, October 17, 2012
   1  
   

   View Slide

2. Insights for the web’s best publishers
   

   View Slide

3. What makes Dash different?
   Dash is purpose-built for publishers and media
   companies. We believe insight > data. Our simple,
   elegant, and intuitive interface requires no training.
   And your tech team will love our easy integration.
   Simply put, you get the best insights to increase
   audience and engagement on your site.
   

   View Slide

4. 4  
   Join the web’s best publishers
   “the best part of working with Parse.ly is
   just how good you have been about
   implementation ... It looks like you guys
   have a more agile development
   environment...”
   “[It] gave us insight into the content
   and our performance explained in
   publisher language rather than your
   standards analytics jargon.”
   Jason Marlin,
   Director of Technology, ArsTechnica
   Zee Kane
   Editor-in-Chief, The Next Web
   

   View Slide

5. View Slide

6. Parse.ly  Mission  
   •  Empower  editors  &  writers  
   •  Liberate  product  teams  
   •  Catalyze  adop>on  of  seman>c  web  
   •  Help  with  sustainability  
   6  
   

   View Slide

7. 7  
   Empower  editors,  writers,  &  analysts  
   with  !mely,  relevant,  &  ac!onable  
   engagement  data  about  web  content.  
   

   View Slide

8. 8  
   Liberate  product  teams  from  the  CMS,  
   with  powerful  data-­‐driven  APIs  to  assist  
   virtuous  user  behaviors.  
   

   View Slide

9. 9  
   Catalyze  adop>on  of  seman!c  web  
   standards  across  the  media  industry.  
   

   View Slide

10. 10  
    Help  media  companies  achieve  
    profitability  &  sustainability  in  digital.  
    

    View Slide

11. 11  
    Scale.  
    

    View Slide

12. View Slide

13. What  kind  of  scale?  
    •  >3  billion  pageviews  per  month  
    •  >10  million  crawled  ar!cles  
    •  >2,500  requests  per  second  at  peak  
    •  ~70  server  nodes  across  three  data  centers  
    •  >Terabyte  of  RAM  with  produc>on  data  
    13  
    

    View Slide

14. 14  
    Data.  
    

    View Slide

15. View Slide

16. 16  
    d3.js  –  Data-­‐Driven  Documents  
    

    View Slide

17. View Slide

18. View Slide

19. 19  
    Metadata.  
    

    View Slide

20. 20  
    rNews
    

    View Slide

21. 21  
    Standard! Implementation! Primary Purpose! Coverage!
    OpenGraph! Multiple META tags" Facebook Rich Embeds" ~60%"
    Schema.org Article! Microdata" SEO" ~80%"
    hNews! Microdata" News Industry Standard" ~90%"
    rNews! RDFa & Microdata" News Industry Standard" ~100%"
    HTML5! Tags" W3C Standard" ~20%"
    parsely-page! Single META tag" Semantic Analytics" ~70%"
    

    View Slide

22. Field! OpenGraph! rNews! HTML5! parsely-page!
    Title! og:title" headline" " title"
    Pub Date! a:published_time" datePublished" " pub_date"
    Body! N/A" articleBody" " N/A*"
    Author! a:author" creator" rel=“author”" author"
    Section! a:section" articleSection" N/A" section"
    Tags! a:tag" about" N/A" tags"
    Page Type! og:type" N/A" N/A" type"
    Canonical Link! og:url" N/A" rel=“canonical”" link"
    Post ID! N/A" identifier" N/A" post_id"
    Main Image! og:image" associatedMedia" N/A" image_url"
    

    View Slide

23. View Slide

24. 24  
    schema.to  
    •  Meet  Mr.  Schemato!  A  friendly  seman>c  web  
    robot  that  makes  metadata  cool  again.  
    •  Open  source  
    •  Public  service  
    •  Eases  implementa!on  with  validators  
    •  Eases  consump!on  with  normalizers  
    •  Extensible  
    

    View Slide

25. 25  
    HTML5,  hNews,  rNews,  
    Schema.org,  OpenGraph,  
    parsely-­‐page  
    {!
    “implements”: {!
    “ogp”: true, !
    “rnews”: true!
    },!
    “distilled”: {!
    “title”: “The Bookstore’s Last Stand”!
    “link”: “http://nytimes.com/123/…”!
    “pub_date”: “2012-01-28”,!
    “image_url”: “http://img.nyt.com/…”,!
    “author”: “Julie Bosman”,!
    “section”: “Business Day”,!
    “tags”: “Barnes & Noble”, “Amazon”,!
    “type”: “post”,!
    “post_id”: “100000001318096”!
    },!
    “extracted”: {...}!
    }!
    

    View Slide

26. 26  
    Here  Today   Coming  Soon  
    •  Valida>on  Framework  
    •  Dis>lling  Framework  
    •  Web  Service  
    •  Validators:  
    –  rNews  
    –  Schema.org  
    –  OpenGraph  
    –  parsely-­‐page  
    •  Dis>llers  
    •  Site  Registry  
    •  Proxy  
    •  Command-­‐Line  Tools  
    •  Validators:  
    –  hNews  
    –  Dublin  Core  
    –  HTML5  
    

    View Slide

27. Parse.ly  Mission  
    •  Empower  editors  &  writers  
    •  Liberate  product  teams  
    •  Catalyze  adop>on  of  seman>c  web  
    •  Help  with  sustainability  
    27  
    

    View Slide

28. 28  
    1 line of javascript that will not
    break or slow down your site
    Streamlined Integration
    Publishers sign up for a free 30-day trial at:
    http://dash.to/try
    

    View Slide

29. Get  in  Touch  
    •  Tweet  us  (now!)  
    – @amontalen>  
    – @parsely  
    •  Email  us  (whenever!)  
    – andrew@parsely.com  
    – hello@parsely.com  
    Open  source  contribu>ons  to  Schema.to,  
    Parse.ly  demos,  or  anything  else!  
    29  
    

    View Slide