1 / 15

CS345 Data Mining

CS345 Data Mining. Virtual Databases. Virtual Relations. Web. Example. Find marketing manager openings in Internet companies so that my commute is shorter than 10 miles. . Structured queries e.g., in SQL. Applications. Comparison shopping shopping.com, fatlens, mobissimo,… Job search

sondra
Download Presentation

CS345 Data Mining

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. CS345Data Mining Virtual Databases

  2. Virtual Relations Web Example • Find marketing manager openings in Internet companies so that my commute is shorter than 10 miles. Structured queries e.g., in SQL

  3. Applications • Comparison shopping • shopping.com, fatlens, mobissimo,… • Job search • indeed.com, simplyhired,… • Classifieds Search • oodle • Integrating web data with relational enterprise apps • purchasing, pricing,…

  4. Website Wrappers • Extract tuples from a single website • Assume website is a static collection of pages i.e., no forms Wrapper Relation

  5. Not same as Relation Extraction • Why can’t we use DIPRE or Snowball? • Can’t assume that the same tuple can be found on many different websites • Need to extract all the tuples from each website • May need to normalize data values across websites • Data may be behind forms • Need to account for query capabilities of websites

  6. Brute force approach • Write a custom program tailored to the website • e.g., in perl, python,… • Does not scale to thousands of websites • Each site needs a different wrapper • Website changes break wrappers

  7. Simpler problem • Simplified version of wrapper problem • Given a set of pages from the same website, that share the same structure • E.g., product detail pages from Amazon.com • We have a target relation schema • E.g., (product,description,price) • Human labels a small subset of pages • Marks tuple components on pages • Can we deduce the structure?

  8. Two web pages <body><h1>Apple 20GB iPod</h1> <img href=“xyz”> Our Price: $204.99 <p> Cool product. </body> <body><h1>Apple 4GB iPod nano</h1> <img href=“abc”> Our Price: $250.99 <p> Even cooler product. </body>

  9. Labeled pages <body><h1>Apple 20GB iPod</h1> <img href=“xyz”> Our Price: $204.99 <p> Cool product. </body> <body><h1>Apple 4GB iPod nano</h1> <img href=“abc”> Our Price: $250.99 <p> Even cooler product. </body>

  10. LR (Left-Right) Wrapper L1 = “<body><h1>” R1 = “</h1><img href=” L2 = “Our Price: ” R2 = “<p>” L3 = “<p>” R3 = “</body>” <body><h1>Apple 20GB iPod</h1> <img href=“xyz”> Our Price: $204.99 <p> Cool product. </body> • Fix an order for attributes (product, price, description) • Use patterns of the form *Li(attributei)Ri*

  11. Example: (Product, Price) <body> <b>Holiday Sale</b><em>save $$</em> <p> <b>Shoes:</b><em>$100</em> <br> <b>Ship:</b><em>$1000</em> </body> <body> <b>Everyday low prices</b><em>guaranteed</em> <p> <b>Sealing wax:</b><em>$1</em> </body> L1=“<b>” R1=“</b>” L2=“<em>” R2=“</em>”

  12. HLRT (Head-Left-Right-Tail) Wrappers <body> <b>Holiday Sale</b><em>save $$</em> <p> <b>Shoes:</b><em>$100</em> <br> <b>Ship:</b><em>$1000</em> </body> <body> <b>Everyday low prices</b><em>guaranteed</em> <p> <b>Sealing wax:</b><em>$1</em> </body> L1=“<b>” R1=“</b>” L2=“<em>” R2=“</em>” H=“</em><p>” T = “<body>”

  13. Example: (Product, Price) <body> <b>Holiday Sale</b><em>save $$</em> <p> <b>Shoes:</b><em>$100</em> <br> <b>Ship:</b><em>$1000</em> </body> <body> <b>Cabbages</b><em>$10</em> <p> <b>Sealing wax:</b><em>$1</em> </body> Cannot construct a HLRT wrapper

  14. Book-author-year example Books by <b>Isaac Asimov</b><ul> <li>Foundation (1951)</li> <li>Nightfall (1941)</li> </ul><p> Books by <b>Arthur C Clarke</b><ul> <li>Rendezvous with Rama (1976)</li> </ul>

  15. Limitations of HLRT • Contiguous tuples • All tuple components must be on the same page • One tuple must end before next one begins • Needs human labeling • Because labeling needs to be accurate • Can we use “noisy” automatic taggers that can make some mistakes?

More Related