1. I256 Applied Natural Language Processing Fall 2009 Lecture 13 Information Extraction (1) Barbara Rosario

2. 2 Today Another project proposal Classification recap Information Extraction (1) Midterm evaluations

3. 3 Classifying at Different Granularities Text Categorization: –Classify an entire document Information Extraction (IE): –Identify and classify small units within documents Named Entity Extraction (NE): –A subset of IE –Identify and classify proper names People, locations, organizations

4. 4 What is Information Extraction? Filling slots in a database from unstructured text. As a task: October 14, 2002, 4:00 a.m. PT For years, Microsoft Corporation CEO Bill Gates railed against the economic philosophy of open-source software with Orwellian fervor, denouncing its communal licensing as a "cancer" that stifled technological innovation. Today, Microsoft claims to "love" the open- source concept, by which software code is [..]. "We can be open source. We love the concept of shared source," said Bill Veghte, a Microsoft VP. "That's a super-important shift for us in terms of code access.“ Richard Stallman, founder of the Free Software Foundation, countered saying… NAME TITLE ORGANIZATION Adapted from slide by William Cohen

5. 5 Information Extraction Extract structured data, such as tables, from unstructured text Task: October 14, 2002, 4:00 a.m. PT For years, Microsoft Corporation CEO Bill Gates railed against the economic philosophy of open-source software with Orwellian fervor, denouncing its communal licensing as a "cancer" that stifled technological innovation. Today, Microsoft claims to "love" the open- source concept, by which software code is [..]. "We can be open source. We love the concept of shared source," said Bill Veghte, a Microsoft VP. "That's a super-important shift for us in terms of code access.“ Richard Stallman, founder of the Free Software Foundation, countered saying… NAME TITLE ORGANIZATION Bill Gates CEO Microsoft Bill Veghte VP Microsoft Richard Stallman founder Free Soft.. IE Adapted from slide by William Cohen

6. 6 Information Extraction Information Extraction = segmentation + classification + association As a family of techniques: October 14, 2002, 4:00 a.m. PT For years, Microsoft Corporation CEO Bill Gates railed against the economic philosophy of open-source software with Orwellian fervor, denouncing its communal licensing as a "cancer" that stifled technological innovation. Today, Microsoft claims to "love" the open- source concept, by which software code is made public to encourage improvement and development by outside programmers. Gates himself says Microsoft will gladly disclose its crown jewels--the coveted code behind the Windows operating system--to select customers. "We can be open source. We love the concept of shared source," said Bill Veghte, a Microsoft VP. "That's a super-important shift for us in terms of code access.“ Richard Stallman, founder of the Free Software Foundation, countered saying… Microsoft Corporation CEO Bill Gates Microsoft Gates Microsoft Bill Veghte Microsoft VP Richard Stallman founder Free Software Foundation aka “named entity extraction/detection” Adapted from slide by William Cohen

7. 7 Information Extraction Information Extraction = segmentation + classification + association A family of techniques: October 14, 2002, 4:00 a.m. PT For years, Microsoft Corporation CEO Bill Gates railed against the economic philosophy of open-source software with Orwellian fervor, denouncing its communal licensing as a "cancer" that stifled technological innovation. Today, Microsoft claims to "love" the open- source concept, by which software code is made public to encourage improvement and development by outside programmers. Gates himself says Microsoft will gladly disclose its crown jewels--the coveted code behind the Windows operating system--to select customers. "We can be open source. We love the concept of shared source," said Bill Veghte, a Microsoft VP. "That's a super-important shift for us in terms of code access.“ Richard Stallman, founder of the Free Software Foundation, countered saying… Microsoft Corporation CEO Bill Gates Microsoft Gates Microsoft Bill Veghte Microsoft VP Richard Stallman founder Free Software Foundation Adapted from slide by William Cohen

8. 8 Information Extraction Information Extraction = segmentation + classification + association A family of techniques: Microsoft Corporation CEO Bill Gates Microsoft Gates Microsoft Bill Veghte Microsoft VP Richard Stallman founder Free Software Foundation Adapted from slide by William Cohen October 14, 2002, 4:00 a.m. PT For years, Microsoft Corporation CEO Bill Gates railed against the economic philosophy of open-source software with Orwellian fervor, denouncing its communal licensing as a "cancer" that stifled technological innovation. Today, Microsoft claims to "love" the open- source concept, by which software code is made public to encourage improvement and development by outside programmers. Gates himself says Microsoft will gladly disclose its crown jewels--the coveted code behind the Windows operating system--to select customers. "We can be open source. We love the concept of shared source," said Bill Veghte, a Microsoft VP. "That's a super-important shift for us in terms of code access.“ Richard Stallman, founder of the Free Software Foundation, countered saying…

9. 9 Landscape of IE Tasks: Degree of Formatting Grammatical sentences and some formatting & links Text paragraphs without formatting Astro Teller is the CEO and co-founder of BodyMedia. Astro holds a Ph.D. in Artificial Intelligence from Carnegie Mellon University, where he was inducted as a national Hertz fellow. His M.S. in symbolic and heuristic computation and B.S. in computer science are from Stanford University. His work in science, literature and business has appeared in international media from the New York Times to CNN to NPR. Non-grammatical snippets, rich formatting & links Tables

10. 10 Landscape of IE Tasks: Intended Breadth of Coverage Web site specificGenre specificWide, non-specific Amazon.com Book PagesResumesUniversity Names FormattingLayoutLanguage Adapted from slide by William Cohen

11. 11 Landscape of IE Tasks: Complexity Closed set He was born in Alabama… The big Wyoming sky… U.S. states Regular set Phone: (413) 545-1323 The CALD main office can be reached at 412-268-1299 U.S. phone numbers Complex pattern University of Arkansas P.O. Box 140 Hope, AR 71802 U.S. postal addresses Headquarters: 1128 Main Street, 4th Floor Cincinnati, Ohio 45210 …was among the six houses sold by Hope Feldman that year. Ambiguous patterns, needing context and many sources of evidence Person names Pawel Opalinski, Software Engineer at WhizBang Labs. Adapted from slide by William Cohen

12. 12 Applications Information Extraction’s applications –Extract structured data out of electronically-available scientific literature, especially in the domain of biology and medicine –Legal documents –Business intelligence –Resume harvesting –Media analysis –Sentiment detection –Patent search –Email scanning

13. 13 Information Extraction Architecture

14. 14 Main tasks Named Entity Recognition Relation Extraction Relations like subject are syntactic, relations like person, location, agent or message are semantic

15. 15 Landscape of IE Tasks Jack Welch will retire as CEO of General Electric tomorrow. The top role at the Connecticut company will be filled by Jeffrey Immelt. Binary relationship Relation Extraction Relation: Person-Title Person: Jack Welch Title: CEO Relation: Company-Location Company: General Electric Location: Connecticut N-ary record Relation: Succession Company: General Electric Title: CEO Out: Jack Welsh In: Jeffrey Immelt Single entity Named entity recognition Person: Jack Welch Person: Jeffrey Immelt Location: Connecticut Adapted from slide by William Cohen

16. 16 Named entity recognition The goal of a named entity recognition (NER) system is to identify all textual mentions of the named entities. Named entities: definite noun phrases that refer to specific types of individuals, such as organizations, persons, dates etc.. Can be broken down into two sub-tasks: 1.identifying the boundaries of the NE (segmentation) 2.identifying its type (classification) Jack Welch will retire as CEO of General Electric tomorrow. The top role at the Connecticut company will be filled by Jeffrey Immelt. Identification of NE

17. 17 Named entity recognition The goal of a named entity recognition (NER) system is to identify all textual mentions of the named entities. Named entities: definite noun phrases that refer to specific types of individuals, such as organizations, persons, dates etc.. Can be broken down into two sub-tasks: 1.identifying the boundaries of the NE (segmentation) 2.identifying its type (classification) Jack Welch will retire as CEO of General Electric tomorrow. The top role at the Connecticut company will be filled by Jeffrey Immelt. Classification of the NE

18. 18 Methods Look up each word in an appropriate list of names. For example, for locations, we could use a gazetteer, or geographical dictionary, such as the Alexandria Gazetteer or the Getty Gazetteer. –error-prone; case distinctions may help, but these are not always present. Location Detection by Simple Lookup for a News Story:

19. 19 Ambiguity Many named entity terms are ambiguous. –May and North are likely to be parts of named entities for DATE and LOCATION, respectively, but could both be part of a PERSON –Christian Dior looks like a PERSON but is more likely to be of type ORGANIZATION. –A term like Yankee will be ordinary modifier in some contexts, but will be marked as an entity of type ORGANIZATION in the phrase Yankee infielders. Further challenges: –Multi-word names like Stanford University –Names that contain other names such as Cecil H. Green Library and Escondido Village Conference Service Center. In named entity recognition, therefore, we need to be able to identify the beginning and end of multi-token sequences  chunking

20. 20 Chunking Chunking useful for entity recognition Segment and label multi-token sequences Each of these larger boxes is called a chunk

21. 21 Chunking The CoNLL 2000 corpus contains 270k words of Wall Street Journal text, annotated with part-of- speech tags and chunk tags. Three chunk types in CoNLL 2000: NP chunks VP chunks PP chunks

22. 22 Chunking More info in Section 7.2 NLTK book We may cover this later during the course but for now, just remember that: You probably need to do chunking if you do a named entity recognition task And perhaps also parse-tree-based features:

23. 23 Path Features From Dan Kein’s CS 288 slides (UC Berkeley)

24. 24 NLTK NE classifier NLTK provides a classifier that has already been trained to recognize named entities, accessed with the function nltk.ne_chunk(). –If we set the parameter binary=True, then named entities are just tagged as NE; otherwise, the classifier adds category labels such as PERSON, ORGANIZATION, and GPE (geo-political entity).

25. 25 NE methods Next class

26. 26 Relation Extraction Once named entities have been identified in a text, we then want to extract the relations that exist between them –Typically relations between specified types of named entity. –Example: look for all triples of the form (X, α, Y), where X and Y are named entities of the required types, and α is the string of words that intervenes between X and Y.

27. 27 Relation Extraction Example: RE: hard-code patterns that contain strings that express the relation that we are looking for. –Example: look for all triples of the form (X, α, Y), where X and Y are named entities of the required types, and α is the string of words that intervenes between X and Y.

28. 28 Relation Extraction nltk.sem.extract_rels Problem with hand-crafted patterns? Machine-learning systems which typically attempt to learn ‘relation patterns’ automatically from a training corpus. Next week we’ll see some machine learning methods to tackle this

29. 29 Semantic Parsers Jack Welch will retire as CEO of General Electric tomorrow. The top role at the Connecticut company will be filled by Jeffrey Immelt. Binary relationship Relation Extraction Relation: Person-Title Person: Jack Welch Title: CEO Relation: Company-Location Company: General Electric Location: Connecticut N-ary record Semantic Parsers Relation: Succession Company: General Electric Title: CEO Out: Jack Welsh In: Jeffrey Immelt Single entity Named entity recognition Person: Jack Welch Person: Jeffrey Immelt Location: Connecticut Adapted from slide by William Cohen

30. 30 PropBank / FrameNet FrameNet: roles shared between verbs PropBank: each verb has it’s own roles PropBank more used, because it’s layered over the treebank (and so has greater coverage, plus parses) Note: some linguistic theories postulate even fewer roles than FrameNet (e.g. 5-20 total: agent, patient, instrument, etc.) From Dan Kein’s CS 288 slides (UC Berkeley)

31. 31 PropBank Example From Dan Kein’s CS 288 slides (UC Berkeley)

32. 32 PropBank Example From Dan Kein’s CS 288 slides (UC Berkeley)

33. 33 PropBank Example From Dan Kein’s CS 288 slides (UC Berkeley)

34. 34 Message Understanding Conference (MUC) DARPA funded significant efforts in IE in the early to mid 1990’s. Message Understanding Conference (MUC) was an annual event/competition where results were presented. Focused on extracting information from news articles: –Terrorist events –Industrial joint ventures –Company management changes Information extraction of particular interest to the intelligence community (CIA, NSA).

35. 35 Message Understanding Conference (MUC) Named entity Person, Organization, Location Co-reference Clinton  President Bill Clinton Template element Perpetrator, Target Template relation Incident Multilingual

36. 36 MUC Typical Text Bridgestone Sports Co. said Friday it has set up a joint venture in Taiwan with a local concern and a Japanese trading house to produce golf clubs to be shipped to Japan. The joint venture, Bridgestone Sports Taiwan Co., capitalized at 20 million new Taiwan dollars, will start production of 20,000 iron and “metal wood” clubs a month

37. 37 MUC Typical Text Bridgestone Sports Co. said Friday it has set up a joint venture in Taiwan with a local concern and a Japanese trading house to produce golf clubs to be shipped to Japan. The joint venture, Bridgestone Sports Taiwan Co., capitalized at 20 million new Taiwan dollars, will start production of 20,000 iron and “metal wood” clubs a month

38. 38 Bridgestone Sports Co. said Friday it had set up a joint venture in Taiwan with a local concern and a Japanese trading house to produce golf clubs to be supplied to Japan. The joint venture, Bridgestone Sports Taiwan Co., capitalized at 20 million new Taiwan dollars, will start production in January 1990 with production of 20,000 iron and “metal wood” clubs a month. Example of IE from FASTUS (1993) TIE-UP-1 MUC template Relationship: TIE-UP Entities: “Bridgestone Sport Co.” “a local concern” “a Japanese trading house” Joint Venture Company: “Bridgestone Sports Taiwan Co.” Activity: ACTIVITY-1 Amount: NT$200000000

39. 39 Bridgestone Sports Co. said Friday it had set up a joint venture in Taiwan with a local concern and a Japanese trading house to produce golf clubs to be supplied to Japan. The joint venture, Bridgestone Sports Taiwan Co., capitalized at 20 million new Taiwan dollars, will start production in January 1990 with production of 20,000 iron and “metal wood” clubs a month. Example of IE: FASTUS(1993) TIE-UP-1 MUC template Relationship: TIE-UP Entities: “Bridgestone Sport Co.” “a local concern” “a Japanese trading house” Joint Venture Company: “Bridgestone Sports Taiwan Co.” Activity: ACTIVITY-1 Amount: NT$200000000 ACTIVITY-1 MUC template Activity: PRODUCTION Company: “Bridgestone Sports Taiwan Co.” Product: “iron and ‘metal wood’ clubs” Start Date: DURING: January 1990

40. 40 Three generations of IE systems Hand-Built Systems – Knowledge Engineering [1980s– ] –Rules written by hand –Require experts who understand both the systems and the domain –Iterative guess-test-tweak-repeat cycle Automatic, Trainable Rule-Extraction Systems [1990s– ] –Rules discovered automatically using predefined templates, using automated rule learners Statistical Models [1997 – ] –Use machine learning to learn which features indicate boundaries and types of entities. –Learning usually supervised; may be partially unsupervised

41. 41 Successors to MUC CoNNL: Conference on Computational Natural Language Learning –Different topics each year –2002, 2003: Language-independent NER –2004: Semantic Role recognition –2001: Identify clauses in text –2000: Chunking boundaries http://cnts.uia.ac.be/conll2003/ (also conll2004, conll2002…)http://cnts.uia.ac.be/conll2003/ ACE: Automated Content Extraction –Entity Detection and Tracking Sponsored by NIST http://wave.ldc.upenn.edu/Projects/ACE/ Several others recently –See http://cnts.uia.ac.be/conll2003/ner/http://cnts.uia.ac.be/conll2003/ner/

42. 42 CoNNL-2003 Goal: identify boundaries and types of named entities –People, Organizations, Locations, Misc. –Experiment with incorporating external resources (Gazeteers) and unlabeled data Data: 4 pieces of info for each term Word POS Chunk EntityType

43. 43 Summary of Results 16 systems participated Machine Learning Techniques –Combinations of Maximum Entropy Models (5) + Hidden Markov Models (4) + Winnow/Perceptron (4) –Others used once were Support Vector Machines, Conditional Random Fields, Transformation-Based learning, AdaBoost, and memory-based learning –Combining techniques often worked well Features –Choice of features is at least as important as ML method –Top-scoring systems used many types –No one feature stands out as essential (other than words) Sang and De Meulder, Introduction to the CoNLL-2003 Shared Task: Language-Independent Named Entity Recognition, Proceedings of CoNLL-2003

44. 44 Evaluating IE Accuracy Always evaluate performance on independent, manually-annotated test data not used during system development. Measure for each test document: –Total number of correct extractions in the solution template: N –Total number of slot/value pairs extracted by the system: E –Number of extracted slot/value pairs that are correct (i.e. in the solution template): C Compute average value of metrics adapted from IR: –Recall = C/N –Precision = C/E –F-Measure = Harmonic mean of recall and precision

45. 45 Sang and De Meulder, Introduction to the CoNLL-2003 Shared Task: Language-Independent Named Entity Recognition, Proceedings of CoNLL-2003

46. 46 Use of External Information Improvement from using Gazeteers vs. unlabeled data nearly equal Gazeteers less useful for German than English (higher quality) Note: standard methods to understand the impact of some features (for a given algorithm); remove them and see error reductions

47. 47 Precision, Recall, and F-Scores * * * * * * Not significantly different Sang and De Meulder, Introduction to the CoNLL-2003 Shared Task: Language-Independent Named Entity Recognition, Proceedings of CoNLL-2003

48. 48 Combining Results What happens if we combine the results of all of the systems? –Used a majority-vote of 5 systems for each set –English: F = 90.30 (14% error reduction of best system) –German: F = 74.17 (6% error reduction of best system)

49. 49 IE Tools Research tools –Gate http://gate.ac.uk/ –MinorThird http://minorthird.sourceforge.net/ –Alembic (only NE tagging) http://www.mitre.org/tech/alembic-workbench/ Commercial –?? I don’t know which ones work well

50. 50 Resources Not checked but from http://nlp.stanford.edu/links/statnlp.htmlhttp://nlp.stanford.edu/links/statnlp.html Semantic Parsers ASSERT –PropBank semantic roles (and opinions, etc.) by Sameer Pradhan. Shalmaneser –FrameNet-based by Katrin Erk. Tree Kernels in SVMlight by Alessandro Moschitti.Tree Kernels in SVMlight –A general package, but it has particularly been used for SRL. Named Entity Recognition Stanford Named Entity Recognizer –A Java Conditional Random Field sequence model with trained models for Named Entity Recognition. Java. GPL. By Jenny Finkel. LingPipe –Tools include statistical named-entity recognition, a heuristic sentence boundary detector, and a heuristic within-document coreference resolution engine. Java. GPL. By Bob Carpenter, Breck Baldwin and co. YamCha –SVM-based NP-chunker, also usable for POS tagging, NER, etc. C/C++ open source. Won CoNLL 2000 shared task. (Less automatic than a specialized POS tagger for an end user.)

51. 51 Resources Information Extraction/Wrapper Induction –Wrapper induction is a method for automatically constructing "wrappers", or scripts which automate the process of retrieving information from a lightly- structured information resource. –Introduction to Information Extraction Technology. A tutorial by Douglas E. Appelt and David Israel.Introduction to Information Extraction Technology –IE data setsIE data sets –Updated versions (i.e., now well-formed XML) of classic IE data sets: Seminar Announcements and Corporate Acquisitions. –Web -> KB. CMU World Wide Knowledge Base project (Tom Mitchell). Has a lot of the best recent probabilistic model IE work, and links to data sets.Web -> KB –RISE: Repository of Online Information Sources Used in Information Extraction Tasks, including links to people, papers, and many widely used data sets, etc. (Ion Muslea). Appears to not have been updated since 1999.RISE –Web IR and IE (Einat Amitay). Various links on IR and IE on the web.Web IR and IE –Web question answering system (University of Michigan)Web question answering system –GATE: General Architecture for Text Engineering (Sheffield)GATE –Genia Project. Biomedical text information extraction corpus (Tsujii lab). And IE tutorial slides.Genia Project Not checked but from http://nlp.stanford.edu/links/statnlp.htmlhttp://nlp.stanford.edu/links/statnlp.html