Information Extraction: Distilling Structured Data from Unstructured Text.<br />Presenter: Shanshan Lu<br />03/04/2010<br />
Referenced paper<br />Andrew McCallum: Information Extraction: Distilling Structured Data from Unstructured Text. ACM Queu...
Example <br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Task: try to build a website ...
Information Extraction: Distilling Structured Data from Unstructured Text<br />
Information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Information ext...
Information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Information ext...
Technique in information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />So...
A Machine Learning Approach<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />A ...
Contributions <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />The ability to ...
Learning extraction rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />One ...
Samples <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />
Rules <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Start rules: <br />End ...
STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />STALKER...
STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />For ins...
Then use another rule to break the list into tuples that correspond to individual restaurants;
finally, from each such tuple they extract the name, address, and phone number of the corresponding restaurant.</li></li><...
STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learnin...
STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learnin...
STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learnin...
Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Appr...
Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Appr...
Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Appr...
Identifying highly informative examples<br />We applied co-testing on the 24 tasks on which STALKER fails to learn perfect...
Verifying the extracted data<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />S...
Automatically repairing wrappers<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br...
How to locate the correct example?<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<...
Automatically repairing wrappers<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />
Upcoming trends and capabilities<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Combi...
Upcoming trends and capabilities<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />SQL -...
Information extraction, the Web and the future<br />Information Extraction: Distilling Structured Data from Unstructured T...
Information extraction, the Web and the future<br />Information Extraction: Distilling Structured Data from Unstructured T...
Upcoming SlideShare
Loading in...5
×

Information Extraction

771

Published on

0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
771
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
34
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Information Extraction

  1. 1. Information Extraction: Distilling Structured Data from Unstructured Text.<br />Presenter: Shanshan Lu<br />03/04/2010<br />
  2. 2. Referenced paper<br />Andrew McCallum: Information Extraction: Distilling Structured Data from Unstructured Text. ACM Queue, volume 3, Number 9, November 2005. <br />Craig A. Knoblock, Kristina Lerman, Steven Minton, Ion Muslea: Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach. IEEE Data Eng. Bull. 23(4): 33-41 (2000)<br />
  3. 3. Example <br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Task: try to build a website to help people find continuing education opportunities at colleges, universities, and organization across the country, to support field searches over locations, dates, times etc.<br />Problem: much of the data was not available in structured form.<br />The only universally available public interfaces were web pages designed for human browsing.<br />
  4. 4. Information Extraction: Distilling Structured Data from Unstructured Text<br />
  5. 5. Information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Information extraction is the process of filling the fields and records of a database from unstructured or loosely formatted text. <br />
  6. 6.
  7. 7. Information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Information extraction involves five major subtasks <br />
  8. 8. Technique in information extraction<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Some simple extraction tasks can be solved by writing regular expressions. <br />Due to Frequently change of web pages, the previous method is not sufficient for the information extraction task. <br />Over the past decade there has been a revolution in the use of statistical and machine-learning methods for information extraction.<br />
  9. 9. A Machine Learning Approach<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />A wrapper is a piece of software that enables a semi-structured Web source to be queried as if it were a database.<br />
  10. 10. Contributions <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />The ability to learn highly accurate extraction rules.<br />To verify the wrapper to ensure that the correct data continues to be extracted.<br />To automatically adapt to changes in the sites from which the data is being extracted.<br />
  11. 11. Learning extraction rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />One of the critical problems in building a wrapper is defining a set of extraction rules that precisely define how to locate the information on the page.<br />For any given item to be extracted from a page, one needs an extraction rule to locate both the beginning and end of that item.<br />A key idea underlying our work is that the extraction rules are based on “landmarks” (i.e., groups of consecutive tokens) that enable a wrapper to locate the start and end of the item within the page.<br />
  12. 12. Samples <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />
  13. 13. Rules <br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Start rules: <br />End rules are similar to start rules.<br />Disjunctive rules:<br />
  14. 14. STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />STALKER : a hierarchical wrapper induction algorithm that learns extraction rules based on examples labeled by the user.<br />STALKER only requires no more than 10 examples because of the fixed web page format and the hierarchical structure.<br />STALKER exploits the hierarchical structure of the source to constrain the learning problem.<br />
  15. 15. STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />For instance, instead of using one complex rule that extracts all restaurant names, addresses and phone numbers from a page, they take a hierarchical approach.<br /><ul><li>Apply a rule that extracts the whole list of restaurants;
  16. 16. Then use another rule to break the list into tuples that correspond to individual restaurants;
  17. 17. finally, from each such tuple they extract the name, address, and phone number of the corresponding restaurant.</li></li></ul><li>STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Algorithm to learn each rule<br />STALKER is a sequential covering algorithm that, given the training examples E, tries to learn a minimal number of perfect disjuncts that cover all examples in E.<br />A perfect disjunct is a rule that covers at least one training example and on any example the rule matches, it produces the correct result.<br />
  18. 18. STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learning a start rule for address:<br />First, it selects an example, say E4, to guide the search. <br />Second, it generates a set of initial candidates, which are rules that consist of a single 1-token landmark; these landmarks are chosen so that they match the token that immediately precedes the beginning of the address in the guiding example.<br />
  19. 19. STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learning a start rule for address:<br />Because R6 has a better generalization potential, STALKER selects R6 for further refinements. <br />While refining R6, STALKER creates, among others, the new candidates R7, R8, R9, and R10 shown below.<br />
  20. 20. STALKER to learn rules<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Learning a start rule for address:<br />As R10 works correctly on all four examples, STALKER stops the learning process and returns R10.<br />Result of STALKER:<br />In an empirical evaluation on 28 sources <br />STALKER had to learn 206 extraction rules. <br />They learned 182 perfect rules (100% accurate), and another 18 rules that had an accuracy of at least 90%. In other words, only 3% of the learned rules were less that 90% accurate.<br />
  21. 21. Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />The most informative examples illustrate exceptional cases.<br />They have developed an active learning approach called co-testing that analyzes the set of unlabeled examples to automatically select examples for the user to label.<br />Backward:<br />
  22. 22. Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Basic idea:<br />after the user labels one or two examples, the system learns both a forward and a backward rule. <br />Then it runs both rules on a given set of unlabeled pages. Whenever the rules disagree on an example, the system considers that as an example for the user to label next.<br />Co-testing makes it possible to generate accurate extraction rules with a very small number of labeled examples.<br />
  23. 23. Identifying highly informative examples<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Assume that the initial training set consists of E1 and E2, while E3 and E4 are not labeled. Based on these examples, we learn the rules:<br />
  24. 24. Identifying highly informative examples<br />We applied co-testing on the 24 tasks on which STALKER fails to learn perfect rules.<br />The results were excellent: the average accuracy over all tasks improved from 85.7% to 94.2%.<br />Furthermore, 10 of the learned rules were 100% accurate, while another 11 rules were at least 90% accurate.<br />
  25. 25. Verifying the extracted data<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Since the information for even a single field can vary considerably, the system learns the statistical distribution of the patterns for each field. <br />Wrappers can be verified by comparing the patterns of data returned to the learned statistical distribution.<br />When a significant difference is found, an operator can be notified or we can automatically launch the wrapper repair process.<br />
  26. 26. Automatically repairing wrappers<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Locate correct examples of the data field on new pages.<br />Re-label the new pages automatically.<br />Labeled and re-labeled examples re-run through the STALKER to produce the correct rules for this site.<br />
  27. 27. How to locate the correct example?<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />Each new page is scanned to identify all text segments that begin with one of the starting patterns and end with one of the ending patterns. Those segments, which we call candidates.<br />The candidates are then clustered to identify subgroups that share common features (relative position on the page, adjacent landmarks, and whether it is visible to the user). <br />Each group is then given a score based on how similar it is to the training examples. <br />We expect the highest ranked group to contain the correct examples of the data field.<br />
  28. 28. Automatically repairing wrappers<br />Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach<br />
  29. 29. Upcoming trends and capabilities<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Combine IE and data mining to perform text mining as well as improve the performance of the underlying extraction system. <br />Rules mined from a database extracted from a corpus of texts are used to predict additional information to extract from future documents, thereby improving the recall of IE.<br />
  30. 30. Upcoming trends and capabilities<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />SQL --> Database<br />
  31. 31. Information extraction, the Web and the future<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />Second half internet revolution: <br />machine access to this immense knowledge base<br />
  32. 32. Information extraction, the Web and the future<br />Information Extraction: Distilling Structured Data from Unstructured Text<br />In web search there will be a transition from keyword search on documents to higher-level queries: <br />queries where the search hits will be objects, such as people or companies instead of simply documents; <br />queries that are structured and return information that has been integrated and synthesized from multiple pages; <br />queries that are stated as natural language questions (“Who were the first three female U.S. Senators?”) and answered with succinct responses. <br />
  33. 33. Thank you!<br />Any questions?<br />
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×