1. October 2005CSA3180: Text Processing III1 CSA3180: Natural Language Processing Text Processing 3 – Double Lecture Discovering Word Associations Text Classification TF.IDF Clustering/Data Mining Linear and Non-Linear Classification Binary Classification Multi-Class Classification

2. October 2005CSA3180: Text Processing III2 Introduction Slides partly based on Lectures by Barbara Rosario and Preslav Nakov Classification –Text categorization (and other applications) Various issues regarding classification –Clustering vs. classification, binary vs. multi-way, flat vs. hierarchical classification… Introduce the steps necessary for a classification task –Define classes –Label text –Features –Training and evaluation of a classifier

3. October 2005CSA3180: Text Processing III3 Classification Goal: Assign ‘objects’ from a universe to two or more classes or categories Problem Object Categories Tagging Word POS Sense Disambiguation Word The word’s senses Information retrieval Document Relevant/not relevant Sentiment classification Document Positive/negative Author identification Document Authors Language identification Document Languages Text Classification Document Topics

4. October 2005CSA3180: Text Processing III4 Author Identification They agreed that Mrs. X should only hear of the departure of the family, without being alarmed on the score of the gentleman's conduct; but even this partial communication gave her a great deal of concern, and she bewailed it as exceedingly unlucky that the ladies should happen to go away, just as they were all getting so intimate together. Gas looming through the fog in divers places in the streets, much as the sun may, from the spongey fields, be seen to loom by husbandman and ploughboy. Most of the shops lighted two hours before their time--as the gas seems to know, for it has a haggard and unwilling look. The raw afternoon is rawest, and the dense fog is densest, and the muddy streets are muddiest near that leaden-headed old obstruction, appropriate ornament for the threshold of a leaden-headed old corporation, Temple Bar.

5. October 2005CSA3180: Text Processing III5 Author Identification Jane Austen (1775-1817), Pride and Prejudice Charles Dickens (1812-70), Bleak House

6. October 2005CSA3180: Text Processing III6 Author Identification Federalist papers –77 short essays written in 1787-1788 by Hamilton, Jay and Madison to persuade NY to ratify the US Constitution; published under a pseudonym –The authorships of 12 papers was in dispute (disputed papers) –In 1964 Mosteller and Wallace * solved the problem –They identified 70 function words as good candidates for authorships analysis –Using statistical inference they concluded the author was Madison

7. October 2005CSA3180: Text Processing III7 Author Identification Function Words

8. October 2005CSA3180: Text Processing III8 Author Identification

9. October 2005CSA3180: Text Processing III9 Language Identification Tutti gli esseri umani nascono liberi ed eguali in dignità e diritti. Essi sono dotati di ragione e di coscienza e devono agire gli uni verso gli altri in spirito di fratellanza. Alle Menschen sind frei und gleich an Würde und Rechten geboren. Sie sind mit Vernunft und Gewissen begabt und sollen einander im Geist der Brüderlichkeit begegnen. Universal Declaration of Human RightsUniversal Declaration of Human Rights, UN, in 363 languages

10. October 2005CSA3180: Text Processing III10 Language Identification égaux - French eguali - Italian iguales - Spanish edistämään - Finnish għ - Maltese

11. October 2005CSA3180: Text Processing III11 Text Classification http://news.google.com/ Reuters –Collection of (21,578) newswire documents. –For research purposes: a standard text collection to compare systems and algorithms –135 valid topics categories

12. October 2005CSA3180: Text Processing III12 Reuters Newswire Corpus

13. October 2005CSA3180: Text Processing III13 Reuters Sample 2-MAR-1987 16:51:43.42 livestock hog AMERICAN PORK CONGRESS KICKS OFF TOMORROW CHICAGO, March 2 - The American Pork Congress kicks off tomorrow, March 3, in Indianapolis with 160 of the nations pork producers from 44 member states determining industry positions on a number of issues, according to the National Pork Producers Council, NPPC. Delegates to the three day Congress will be considering 26 resolutions concerning various issues, including the future direction of farm policy and the tax law as it applies to the agriculture sector. The delegates will also debate whether to endorse concepts of a national PRV (pseudorabies virus) control and eradication program, the NPPC said. A large trade show, in conjunction with the congress, will feature the latest in technology in all areas of the industry, the NPPC added. Reuter

14. October 2005CSA3180: Text Processing III14 Classification vs. Clustering Classification assumes labeled data: we know how many classes there are and we have examples for each class (labeled data). Classification is supervised In Clustering we don’t have labeled data; we just assume that there is a natural division in the data and we may not know how many divisions (clusters) there are Clustering is unsupervised

15. October 2005CSA3180: Text Processing III15 Classification Class1 Class2

16. October 2005CSA3180: Text Processing III16 Classification Class1 Class2

17. October 2005CSA3180: Text Processing III17 Classification Class1 Class2

18. October 2005CSA3180: Text Processing III18 Classification Class1 Class2

19. October 2005CSA3180: Text Processing III19 Clustering

20. October 2005CSA3180: Text Processing III20 Clustering

21. October 2005CSA3180: Text Processing III21 Clustering

22. October 2005CSA3180: Text Processing III22 Clustering

23. October 2005CSA3180: Text Processing III23 Clustering

24. October 2005CSA3180: Text Processing III24 Categories (Labels, Classes) Labeling data 2 problems: Decide the possible classes (which ones, how many) –Domain and application dependent –http://news.google.comhttp://news.google.com Label text –Difficult, time consuming, inconsistency between annotators

25. October 2005CSA3180: Text Processing III25 Reuters 2-MAR-1987 16:51:43.42 livestock hog AMERICAN PORK CONGRESS KICKS OFF TOMORROW CHICAGO, March 2 - The American Pork Congress kicks off tomorrow, March 3, in Indianapolis with 160 of the nations pork producers from 44 member states determining industry positions on a number of issues, according to the National Pork Producers Council, NPPC. Delegates to the three day Congress will be considering 26 resolutions concerning various issues, including the future direction of farm policy and the tax law as it applies to the agriculture sector. The delegates will also debate whether to endorse concepts of a national PRV (pseudorabies virus) control and eradication program, the NPPC said. A large trade show, in conjunction with the congress, will feature the latest in technology in all areas of the industry, the NPPC added. Reuter Why not topic = policy ?

26. October 2005CSA3180: Text Processing III26 Binary vs. Multi-Class Classification Binary classification: two classes Multi-way classification: more than two classes Sometime it can be convenient to treat a multi-way problem like a binary one: one class versus all the others, for all classes

27. October 2005CSA3180: Text Processing III27 Flat vs. Hierarchical Classification Flat classification: relations between the classes undetermined Hierarchical classification: hierarchy where each node is the sub-class of its parent’s node

28. October 2005CSA3180: Text Processing III28 Single vs. Multi-Category Classification In single-category text classification each text belongs to exactly one category In multi-category text classification, each text can have zero or more categories

29. October 2005CSA3180: Text Processing III29 LabeledText class in NLTK LabeledText classLabeledText >>> text = "Seven-time Formula One champion Michael Schumacher took on the Shanghai circuit Saturday in qualifying for the first Chinese Grand Prix." >>> label = “sport” >>> labeled_text = LabeledText(text, label) >>> labeled_text.text() “Seven-time Formula One champion Michael Schumacher took on the Shanghai circuit Saturday in qualifying for the first Chinese Grand Prix.” >>> labeled_text.label() “sport”

30. October 2005CSA3180: Text Processing III30 NLTK Classifier Interface classify determines which label is most appropriate for a given text token, and returns a labeled text token with that label.classify labels returns the list of category labels that are used by the classifier.labels >>> token = Token(“The World Health Organization is recommending more importance be attached to the prevention of heart disease and other cardiovascular ailments rather than focusing on treatment.”) >>> my_classifier.classify(token) “The World Health Organization is recommending more importance be attached to the prevention of heart disease and other cardiovascular ailments rather than focusing on treatment.”/ health >>> my_classifier.labels() ("sport", "health", "world",…)

31. October 2005CSA3180: Text Processing III31 Features >>> text = "Seven-time Formula One champion Michael Schumacher took on the Shanghai circuit Saturday in qualifying for the first Chinese Grand Prix." >>> label = “sport” >>> labeled_text = LabeledText(text, label) Here the classification takes as input the whole string What’s the problem with that? What are the features that could be useful for this example?

32. October 2005CSA3180: Text Processing III32 Features Feature: An aspect of the text that is relevant to the task Some typical features –Words present in text –Frequency of words –Capitalization –Are there NE? –WordNet –Others?

33. October 2005CSA3180: Text Processing III33 Features Feature: An aspect of the text that is relevant to the task Feature value: the realization of the feature in the text –Words present in text : Kerry, Schumacher, China… –Frequency of word: Kerry(10), Schumacher(1)… –Are there dates? Yes/no –Are there PERSONS? Yes/no –Are there ORGANIZATIONS? Yes/no –WordNet: Holonyms (China is part of Asia), Synonyms(China, People's Republic of China, mainlan d China)

34. October 2005CSA3180: Text Processing III34 Features Boolean (or Binary) Features Features that generate boolean (binary) values. Boolean features are the simplest and the most common type of feature. –f 1 (text) = 1 if text contain “Kerry” 0 otherwise –f 2 (text) = 1 if text contain PERSON 0 otherwise

35. October 2005CSA3180: Text Processing III35 Features Integer Features Features that generate integer values. Integer features can be used to give classifiers access to more precise information about the text. –f 1 (text) = Number of times text contains “Kerry” –f 2 (text) = Number of times text contains PERSON

36. October 2005CSA3180: Text Processing III36 Features in NLTK Feature Detectors –Features can be defined using feature detector functions, which map LabeledTexts to values –Method: detect, which takes a labeled text, and returns a feature value.detect –>>> def ball(ltext): return (“ball” in ltext.text()) –>>> fdetector = FunctionFeatureDetector(ball) –>>> document1 = "John threw the ball over the fence".split() –>>> fdetector.detect(LabeledText(document1) 1 –>>> document2 = "Mary solved the equation".split() –>>> fdetector.detect(LabeledText(document2) 0

37. October 2005CSA3180: Text Processing III37 Features Linguistic features – Words lowercase? (should we convert to?) normalized? (e.g. “texts”  “text”) – Phrases – Word-level n-grams – Character-level n-grams – Punctuation – Part of Speech Non-linguistic features – document formatting – informative character sequences (e.g. &lt)

38. October 2005CSA3180: Text Processing III38 When do we need Feature Selection? If the algorithm cannot handle all possible features e.g. language identification for 100 languages using all words text classification using n-grams Good features can result in higher accuracy – But! Why feature selection? – What if we just keep all features? Even the unreliable features can be helpful. But we need to weight them: –In the extreme case, the bad features can have a weight of 0 (or very close), which is… a form of feature selection!

39. October 2005CSA3180: Text Processing III39 Why do we need Feature Selection? Not all features are equally good! – Bad features: best to remove Infrequent –unlikely to be be met again –co-occurrence with a class can be due to chance Too frequent –mostly function words Uniform across all categories – Good features: should be kept Co-occur with a particular category Do not co-occur with other categories – The rest: good to keep

40. October 2005CSA3180: Text Processing III40 What types of Feature Selection?  Feature selection reduces the number of features  Usually:  Eliminating features  Weighting features  Normalizing features  Sometimes by transforming parameters  e.g. Latent Semantic Indexing using Singular Value Decomposition  Method may depend on problem type  For classification and filtering, may use information from example documents to guide selection

41. October 2005CSA3180: Text Processing III41 What types of Feature Selection? Task independent methods –Document Frequency (DF) –Term Strength (TS) Task-dependent methods –Information Gain (IG) –Mutual Information (MI) –  2 statistic (CHI) Empirically compared by Yang & Pedersen (1997)

42. October 2005CSA3180: Text Processing III42 Document Frequency (DF) DF: number of documents a term appears in Based on Zipf’s Law Remove the rare terms: (met 1-2 times) – Non-informative – Unreliable – can be just noise – Not influential in the final decision – Unlikely to appear in new documents Plus – Easy to compute – Task independent: do not need to know the classes Minus – Ad hoc criterion – Rare terms can be good discriminators (e.g., in IR) What about the frequent terms? What is a “rare” term?

43. October 2005CSA3180: Text Processing III43 Stop Word Removal Common words from a predefined list – Mostly from closed-class categories: unlikely to have a new word added include: auxiliaries, conjunctions, determiners, prepositions, pronouns, articles – But also some open-class words like numerals Bad discriminators – uniformly spread across all classes – can be safely removed from the vocabulary Is this always a good idea? (e.g. author identification)

44. October 2005CSA3180: Text Processing III44  2 statistic (CHI)  2 statistic (pronounced “kai square”) – The most commonly used method of comparing proportions. – Checks whether there is a relationship between being in one of two groups and a characteristic under study. – Example: Let us measure the dependency between a term t and a category c. the groups would be: –1) the documents from a category c i –2) all other documents the characteristic would be: –“document contains term t”

45. October 2005CSA3180: Text Processing III45  2 statistic (CHI) Is “jaguar” a good predictor for the “auto” class? We want to compare: the observed distribution above; and null hypothesis: that jaguar and auto are independent Term = jaguar Term  jaguar Class = auto2500 Class  auto 39500

46. October 2005CSA3180: Text Processing III46  2 statistic (CHI) Under the null hypothesis: (jaguar and auto – independent): How many co-occurrences of jaguar and auto do we expect? – We would have: P r (j,a) = P r (j)  P r (a) – So, there would be: N  P r (j,a), i.e. N  P r (j)  P r (a) – P r (j) = (2+3)/N; P r (a) = (2+500)/N; N=2+3+500+9500 – Which is: N  (5/N)  (502/N)=2510/N=2510/10005  0.25 Term = jaguar Term  jaguar Class = auto2500 Class  auto 39500

47. October 2005CSA3180: Text Processing III47  2 statistic (CHI) Under the null hypothesis: (jaguar and auto – independent): How many co-occurrences of jaguar and auto do we expect? – We would have: P r (j,a) = P r (j)  P r (a) – So, there would be: N  P r (j,a), i.e. N  P r (j)  P r (a) – P r (j) = (2+3)/N; P r (a) = (2+500)/N; N=2+3+500+9500 – Which is: N  (5/N)  (502/N)=2510/N=2510/10005  0.25 Term = jaguar Term  jaguar Class = auto2(0.25)500(502) Class  auto 3(4.75)9500(9498) expected: f e observed: f o

48. October 2005CSA3180: Text Processing III48  2 statistic (CHI)  2 is interested in (f o – f e ) 2 /f e summed over all table entries: The null hypothesis is rejected with confidence.999, since 12.9 > 10.83 (the value for.999 confidence). Term = jaguar Term  jaguar Class = auto2(0.25)500(502) Class  auto 3(4.75)9500(9498) expected: f e observed: f o

49. October 2005CSA3180: Text Processing III49  2 statistic (CHI) How to use  2 for multiple categories? Compute  2 for each category and then combine: – we can require to discriminate well across all categories, then we need to take the expected value of  2 : or to discriminate well for a single category, then we take the maximum:

50. October 2005CSA3180: Text Processing III50  2 statistic (CHI) Pros – normalized and thus comparable across terms –  2 (t,c) is 0, when t and c are independent – can be compared to  2 distribution, 1 degree of freedom Cons – unreliable for low frequency terms – computationally expensive

51. October 2005CSA3180: Text Processing III51 Term Weighting In the study just shown, terms were (mainly) treated as binary features –If a term occurred in a document, it was assigned 1 –Else 0 Often it us useful to weight the selected features Standard technique: tf.idf

52. October 2005CSA3180: Text Processing III52 TF.IDF Term Weighting TF: term frequency – definition: TF = t ij frequency of term i in document j – purpose: makes the frequent words for the document more important IDF: inverted document frequency – definition: IDF = log(N/n i ) n i : number of documents containing term i N : total number of documents – purpose: makes rare words across documents more important TF.IDF – definition: t ij  log(N/n i )

53. October 2005CSA3180: Text Processing III53 Term Normalization Combine different words into a single representation –Stemming/morphological analysis bought, buy, buys -> buy –General word categories $23.45, 5.30 Yen -> MONEY 1984, 10,000 -> DATE, NUM PERSON ORGANIZATION –(Covered in Information Extraction segment) –Generalize with lexical hierarchies WordNet, MeSH –(Covered later in the semester)

54. October 2005CSA3180: Text Processing III54 Stemming and Lemmatization Purpose: conflate morphological variants of a word to a single index term –Stemming: normalize to a pseudoword e.g. “more” and “morals” become “mor” (Porter stemmer) –Lemmatization: convert to the root form e.g. “more” and “morals” become “more” and “moral” Plus: –vocabulary size reduction –data sparseness reduction Minus: –loses important features (even to_lowercase() can be bad!) –questionable utility (maybe just “-s”, “-ing” and “-ed”?)

55. October 2005CSA3180: Text Processing III55 Practical Approach 1.Feature selection infrequent term removal –infrequent across the whole collection (i.e. DF) –met in a single document most frequent term removal (i.e. stop words) 2.Normalization: 1.Stemming. (often) 2.Word classes (sometimes) 3.Feature weighting: TF.IDF or IDF 4.Dimensionality reduction. (occasionally)

56. October 2005CSA3180: Text Processing III56 Classification Linear versus non linear classification Binary classification –Perceptron –Winnow –Support Vector Machines (SVM) –Kernel Methods (covered in statistics lectures) Multi-Class classification (covered in Statistics Lectures) –Decision Trees –Naïve Bayes –K nearest neighbor

57. October 2005CSA3180: Text Processing III57 Binary Classification Spam filtering (spam, not spam) Customer service message classification (urgent vs. not urgent) Information retrieval (relevant, not relevant) Sentiment classification (positive, negative) Sometime it can be convenient to treat a multi- way problem like a binary one: one class versus all the others, for all classes

58. October 2005CSA3180: Text Processing III58 Binary Classification Given: some data items that belong to a positive (+1 ) or a negative (-1 ) class Task: Train the classifier and predict the class for a new data item Geometrically: find a separator

59. October 2005CSA3180: Text Processing III59 Linear vs. Non-Linear Linearly separable data: if all the data points can be correctly classified by a linear (hyperplanar) decision boundary

60. October 2005CSA3180: Text Processing III60 Linear vs. Non-Linear Linear Decision boundary

61. October 2005CSA3180: Text Processing III61 Linear vs. Non-Linear Class1 Class2 Non-Linearly Separable

62. October 2005CSA3180: Text Processing III62 Linear vs. Non-Linear Non Linear Classifier Class1 Class2

63. October 2005CSA3180: Text Processing III63 Linear vs. Non-Linear Linear or Non linear separable data? –We can find out only empirically Linear algorithms (algorithms that find a linear decision boundary) –When we think the data is linearly separable –Advantages Simpler, less parameters –Disadvantages High dimensional data (like for NLT) is usually not linearly separable –Examples: Perceptron, Winnow, SVM –Note: we can use linear algorithms also for non linear problems (see Kernel methods)

64. October 2005CSA3180: Text Processing III64 Linear vs. Non-Linear Non Linear –When the data is non linearly separable –Advantages More accurate –Disadvantages More complicated, more parameters –Example: Kernel methods Note: the distinction between linear and non linear applies also for multi-class classification (we’ll see this later)

65. October 2005CSA3180: Text Processing III65 Simple Linear Algorithms Perceptron and Winnow algorithm –Linear –Binary classification –Online (process data sequentially, one data point at the time) –Mistake driven –Simple single layer Neural Networks

66. October 2005CSA3180: Text Processing III66 Simple Linear Algorithms Data: {(x i,y i )} i=1...n –x in R d (x is a vector in d-dimensional space)  feature vector –y in {-1,+1}  label (class, category) Question: –Design a linear decision boundary: wx + b (equation of hyperplane) such that the classification rule associated with it has minimal probability of error –classification rule: y = sign(w x + b) which means: if wx + b > 0 then y = +1 if wx + b < 0 then y = -1

67. October 2005CSA3180: Text Processing III67 Simple Linear Algorithms Find a good hyperplane (w,b) in R d+1 that correctly classifies data points as much as possible In online fashion: one data point at the time, update weights as necessary wx + b = 0 Classification Rule: y = sign(wx + b)

68. October 2005CSA3180: Text Processing III68 Perceptron Algorithm Initialize: w 1 = 0 Updating rule For each data point x –If class(x) != decision(x,w) –then w k+1  w k + y i x i k  k + 1 –else w k+1  w k Function decision(x, w) –If wx + b > 0 return +1 –Else return - 1 wkwk 0 +1 w k x + b = 0 w k+1 W k+1 x + b = 0

69. October 2005CSA3180: Text Processing III69 Perceptron Algorithm Online: can adjust to changing target, over time Advantages –Simple and computationally efficient –Guaranteed to learn a linearly separable problem (convergence, global optimum) Limitations –Only linear separations –Only converges for linearly separable data –Not really “efficient with many features”

70. October 2005CSA3180: Text Processing III70 Winnow Algorithm Another online algorithm for learning perceptron weights: f(x) = sign(wx + b) Linear, binary classification Update-rule: again error-driven, but multiplicative (instead of additive)

71. October 2005CSA3180: Text Processing III71 Winnow Algorithm Initialize: w 1 = 0 Updating rule For each data point x –If class(x) != decision(x,w) –then w k+1  w k + y i x i  Perceptron w k+1  w k *exp(y i x i )  Winnow k  k + 1 –else w k+1  w k Function decision(x, w) –If wx + b > 0 return +1 –Else return -1 wkwk 0 +1 w k x + b= 0 w k+1 W k+1 x + b = 0

72. October 2005CSA3180: Text Processing III72 Perceptron vs. Winnow Assume –N available features –only K relevant items, with K<<N Perceptron: number of mistakes: O( K N) Winnow: number of mistakes: O(K log N) Winnow is more robust to high-dimensional feature spaces

73. October 2005CSA3180: Text Processing III73 Perceptron vs. Winnow Perceptron Online: can adjust to changing target, over time Advantages –Simple and computationally efficient –Guaranteed to learn a linearly separable problem Limitations –only linear separations –only converges for linearly separable data –not really “efficient with many features” Winnow Online: can adjust to changing target, over time Advantages –Simple and computationally efficient –Guaranteed to learn a linearly separable problem –Suitable for problems with many irrelevant attributes Limitations –only linear separations –only converges for linearly separable data –not really “efficient with many features” Used in NLP

74. October 2005CSA3180: Text Processing III74 Support Vector Machine (SVM) Large Margin Classifier Linearly separable case Goal: find the hyperplane that maximizes the margin w T x + b = 0 M w T x a + b = 1 w T x b + b = -1 Support vectors

75. October 2005CSA3180: Text Processing III75 Support Vector Machine (SVM) Text classification Hand-writing recognition Computational biology (e.g., micro-array data) Face detection Face expression recognition Time series prediction

76. October 2005CSA3180: Text Processing III76 Classification II Non-linear algorithms Kernel methods Multi-class classification Decision trees Na ï ve Bayes Last topic for today: k Nearest Neighbour

77. October 2005CSA3180: Text Processing III77 k Nearest Neighbour Nearest Neighbor classification rule: to classify a new object, find the object in the training set that is most similar. Then assign the category of this nearest neighbor K Nearest Neighbor (KNN): consult k nearest neighbors. Decision based on the majority category of these neighbors. More robust than k = 1 –Example of similarity measure often used in NLP is cosine similarity

78. October 2005CSA3180: Text Processing III78 1 Nearest Neighbour

79. October 2005CSA3180: Text Processing III79 1 Nearest Neighbour

80. October 2005CSA3180: Text Processing III80 3 Nearest Neighbour

81. October 2005CSA3180: Text Processing III81 3 Nearest Neighbour Assign the category of the majority of the neighbors But this is closer.. We can weight neighbors according to their similarity

82. October 2005CSA3180: Text Processing III82 k Nearest Neighbour Strengths –Robust –Conceptually simple –Often works well –Powerful (arbitrary decision boundaries) Weaknesses –Performance is very dependent on the similarity measure used (and to a lesser extent on the number of neighbors k used) –Finding a good similarity measure can be difficult –Computationally expensive