1. Machine Learning for Information Extraction: An Overview Kamal Nigam Google Pittsburgh With input, slides and suggestions from William Cohen, Andrew McCallum and Ion Muslea
2. Example: A Problem Genomics job Mt. Baker, the school district Baker Hostetler , the company Baker, a job opening
5. Extracting Job Openings from the Web Title: Ice Cream Guru Description: If you dream of cold creamy… Contact: [email_address] Category: Travel/Hospitality Function: Food Services
18. Landscape of ML Techniques for IE: Any of these models can be used to capture words, formatting or both. Classify Candidates Abraham Lincoln was born in Kentucky . Classifier which class? Sliding Window Abraham Lincoln was born in Kentucky. Classifier which class? Try alternate window sizes: Boundary Models Abraham Lincoln was born in Kentucky. Classifier which class? BEGIN END BEGIN END BEGIN Finite State Machines Abraham Lincoln was born in Kentucky. Most likely state sequence? Wrapper Induction <b><i> Abraham Lincoln </i></b> was born in Kentucky. Learn and apply pattern for a website <b> <i> PersonName
20. Information Extraction by Sliding Windows GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
21. Information Extraction by Sliding Windows GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
22. Information Extraction by Sliding Window GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
23. Information Extraction by Sliding Window GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
24. Information Extraction by Sliding Window GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
25.
26.
27. IE by Boundary Detection GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
28. IE by Boundary Detection GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
29. IE by Boundary Detection GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
30. IE by Boundary Detection GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
31. IE by Boundary Detection GRAND CHALLENGES FOR MACHINE LEARNING Jaime Carbonell School of Computer Science Carnegie Mellon University 3:30 pm 7500 Wean Hall Machine learning has evolved from obscurity in the 1970s into a vibrant and popular discipline in artificial intelligence during the 1980s and 1990s. As a result of its success and growth, machine learning is evolving into a collection of related disciplines: inductive concept acquisition, analytic learning in problem solving (e.g. analogy, explanation-based learning), learning theory (e.g. PAC learning), genetic algorithms, connectionist learning, hybrid systems, and so on. CMU UseNet Seminar Announcement
32.
33.
34. BWI: Learning to detect boundaries Field F1 Person Name: 30% Location: 61% Start Time: 98%
37. Hidden Markov Models S t - 1 S t O t S t+1 O t +1 O t - 1 ... ... Finite state model Graphical model Parameters: for all states S={s 1 ,s 2 ,…} Start state probabilities: P(s t ) Transition probabilities: P(s t |s t-1 ) Observation (emission) probabilities: P(o t |s t ) Training: Maximize probability of training observations (w/ prior) HMMs are the standard sequence modeling tool in genomics, music, speech, NLP, … ... transitions observations o 1 o 2 o 3 o 4 o 5 o 6 o 7 o 8 Generates: State sequence Observation sequence Usually a multinomial over atomic, fixed alphabet
38. IE with Hidden Markov Models Yesterday Lawrence Saul spoke this example sentence. Yesterday Lawrence Saul spoke this example sentence. Person name: Lawrence Saul Given a sequence of observations: and a trained HMM: Find the most likely state sequence: (Viterbi) Any words said to be generated by the designated “person name” state extract as a person name:
39.
40. HMM Example: “Nymble” Other examples of HMMs in IE: [Leek ’97; Freitag & McCallum ’99; Seymore et al. 99] Task: Named Entity Extraction Train on 450k words of news wire text. Case Language F1 . Mixed English 93% Upper English 91% Mixed Spanish 90% [Bikel, et al 97] Person Org Other (Five other name classes) start-of-sentence end-of-sentence Transition probabilities Observation probabilities P(s t | s t-1 , o t-1 ) P(o t | s t , s t-1 ) Back-off to: Back-off to: P(s t | s t-1 ) P(s t ) P(o t | s t , o t-1 ) P(o t | s t ) P(o t ) or Results:
41.
42.
43.
44. Conditional Markov Models S t - 1 S t O t S t+1 O t +1 O t - 1 ... ... Generative (traditional HMM) ... transitions observations S t - 1 S t O t S t+1 O t +1 O t - 1 ... ... Conditional ... transitions observations Standard belief propagation: forward-backward procedure. Viterbi and Baum-Welch follow naturally. Maximum Entropy Markov Models [McCallum, Freitag & Pereira, 2000] MaxEnt POS Tagger [Ratnaparkhi, 1996] SNoW-based Markov Model [Punyakanok & Roth, 2000]
45. Exponential Form for “Next State” Function Capture dependency on s t-1 with |S| independent functions, P s t-1 (s t |o t ). Each state contains a “next-state classifier” that, given the next observation, produces a probability of the next state, P s t-1 (s t |o t ). s t-1 s t Recipe: - Labeled data is assigned to transitions. - Train each state’s exponential model by maximum entropy weight feature
46.
47. From HMMs to MEMMs to CRFs HMM MEMM CRF S t-1 S t O t S t+1 O t+1 O t-1 S t-1 S t O t S t+1 O t+1 O t-1 S t-1 S t O t S t+1 O t+1 O t-1 ... ... ... ... ... ... (A special case of MEMMs and CRFs.) Conditional Random Fields (CRFs) [Lafferty, McCallum, Pereira ‘2001]
48. Conditional Random Fields (CRFs) S t S t+1 S t+2 O = O t , O t+1 , O t+2 , O t+3 , O t+4 S t+3 S t+4 Markov on s , conditional dependency on o. Hammersley-Clifford-Besag theorem stipulates that the CRF has this form—an exponential function of the cliques in the graph. Assuming that the dependency structure of the states is tree-shaped (linear chain is a trivial tree), inference can be done by dynamic programming in time O(|o| |S| 2 )—just like HMMs.
49.
50.
51.
52.
Notas del editor
Not only did this support extremely useful and targeted search and data navigation, with highly summarized display
CiteSeer, which we all know and love, is made possible by the automatic extraction of title and author information from research paper headers and references in order to build the reference graph that we use to find related work.
ML… although this is an area where ML has not yet trounced the hand-built systems. In some of the latest evaluations, hand-built shared 1 st place with a ML. Now many companies making a business from IE (from the Web): WasBang, Inxight, Intelliseek, ClearForest.
Technical approach? At CMU I worked w/ HMMs. HMMs: standard tool…, It is a model that makes certain Markov and independence assumptions as depicted in this Graphical Model. Pr certain states are the val of this random variable… = this product of independent pr. Emissions are traditionally atomic entities, like words. Big part of power is the context of finite state: Viterbi We say this is generative model because it has parameters for probability of producing observed emissions (given state)
Both in academia and in industry, FSMs are the dominant method of information extraction. HMM acts as a generative model of a research paper header. It has trad. parameters for… Emissions are multinomial distributions over words Params set to maximize likelihood of emissions in training data. HMMs work better than several alternative methods, but...
1. Prefer richer… one that allows for multiple, arbitrary... In many cases, for example names and other large vocabulary fields, the identity of word alone is not very predictive because you probably have never seen it before. For example…”Wisneiwski” … not city In many other cases, such as FAQ seg, the relevant features are not really the words at all, but attributes of the line or para as a whole… emissions=whole lines of text at a time
Technical approach? At CMU I worked w/ HMMs. HMMs: standard tool…, It is a model that makes certain Markov and independence assumptions as depicted in this Graphical Model. Pr certain states are the val of this random variable… = this product of independent pr. Emissions are traditionally atomic entities, like words. Big part of power is the context of finite state: Viterbi We say this is generative model because it has parameters for probability of producing observed emissions (given state)
04/26/10
MEMMs do have one technical problem: biased towards states with few siblings. No time to explain intricasies now. CRFs = generalization of MEMMs: move normalizer Don’t let all the equations scare you: this slide is easier than it looks. Here is the old, traditional HMM: Prob of state and obs sequence is product of each state trans, and observation emission in the sequence. MEMMs: as we discussed, replaces with conditional, where we no longer generate the obs. We write this as an exponential function Only difference between MEMMs and CRFs is that we move the normalizer outside the product. Inference in MRFs usually requires nasty Gibbs sampling, but due to special structure, clever DP comes to the rescue again.
Graphical model on previous slide was a special case. In general, observations don’t have to be chopped up, and feature functions can ask arbitrary questions about any range of the observation sequence. Not going to get into method for learning lambda weights from training data, but simply a matter of maximum likelihood: This is what trying to maximize, differentiate, and solve with Conjugate Gradient. Again, DP makes it efficient.