INF 141: Data Recovery - PowerPoint PPT Presentation

inf 141 information retrieval n.
Skip this Video
Loading SlideShow in 5 Seconds..
INF 141: Data Recovery PowerPoint Presentation
INF 141: Data Recovery

play fullscreen
1 / 28
Download Presentation

INF 141: Data Recovery

Presentation Transcript

  1. INF 141: Information Retrieval Discussion Session Week 3 – Winter 2010 TA: Sara Javanmardi

  2. Open Source Web Crawlers

  3. Heritrix • Extensible, Web-Scale, Distributed • Internet Archive’s Crawler

  4. Internet Archive • dedicated to building and maintaining a free and openly accessible online digital library, including an archive of the Web.

  5. Nutch • Apache’s Open Source Search Engine • Distributed • Tested with 100M Pages

  6. WebSphinx • 1998-2002 • Single Machine • Lots of Problems (Memory leaks, …) • Reported to be very slow

  7. Crawler4j • Single Machine • Should Easily Scale to 20M Pages • Very Fast • Crawled and Processed the whole English Wikipedia in 10 hours.

  8. Architecture Should be Extremely Fast, otherwise it would be a bottleneck

  9. Docid Server • Key-value pairs are stored in a B+-tree data structure. • Berkeley DB as the storage engine

  10. Berkeley DB • Unlike traditional database systems like MySQL and others, Berkeley DB comes in form of a jar file which is linked to the Java program and runs in the process space of the crawlers. • No need for inter-process communication and waiting for context switch between processes. • You can think of it as a large HashMap: Key Value

  11. Adding a URL to Frontier public static synchronized int getDocID(String URL) { if there is any key-value pair for key = URL return value else docID= lastdocid+1 put (URL, docID) in storage return -docID } We add the URL to the frontier put (docID,URL) in URL - Queue

  12. Things to Know: • Crawler4j only handles duplicate detection in the level of URLs, not in the level of the content. • Frontier can be implemented as Priority Queue .

  13. Why Priority Queue? • Politeness: do not hit a web server too frequently • Freshness: crawl some pages more often than others • E.g., pages (such as News sites) whose content changes often

  14. Assigning Priority • Prioritizer assigns to URL an integer priority between 1 and K • Heuristics for assigning priority • Refresh rate sampled from previous crawls • Application-specific (e.g., “crawl news sites more often”)

  15. Assignment 3: Crawling

  16. Assignment 3 Programming Part • You can do the assignment 3 in groups of 1, 2 or 3. Tiffany Siu James Milewski , Matt Fritz Kevin Boomhouwer Azia Foster James Rose , Sean Tsusaki , Jeff Gaskill Tzu Yang Huang Fiel Guhit ,Sarah Lee Rob Duncan, Ben Kahn, Dan Morgan Qi Zhu (Chess), Zhuomin Wu Lucy Luxiao, Melanie Sun, Norik Davtian Alex Kaiser, Sam Kaufman Nery Chapeton Jason Gahagan Melanie Cheung, Anthony Liu Chad Curtis , Derek Lee, Rakesh Rajput Andrew J. Santa Maria Zack Pelz

  17. Crawling One Digg Category … Initial Seeds

  18. User-agent: * Disallow: /aboutpost Disallow: /addfriends Disallow: /addim Disallow: /addlink Disallow: /ajax Disallow: /api Disallow: /captcha Disallow: /css/remote-skins/ Disallow: /deleteuserim Disallow: /deleteuserlink Disallow: /diginfull ... … … User-agent: Referrer Karma/2.0 Disallow: /

  19. Things To Do • Download the jar file and import it. • Download the dependency libraries and import them. • Download and complete the source code to crawl

  20. Extra Credit Question 1) Extract story id from digg page: <div  class="news-body"id="18765384"> 2) Send to API:

  21. Questions?