October 2005CSA3180: Text Processing III1 CSA3180: Natural Language Processing Text Processing 3 – Double Lecture Discovering Word Associations Text Classification

October 2005 CSA3180: Text Processing III 1

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


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


Classification

Goal: Assign ‘objects’ from a universe to two or more classes or categories

Problem Object CategoriesTagging Word POSSense Disambiguation Word The word’s sensesInformation retrieval Document Relevant/not relevantSentiment classification Document Positive/negativeAuthor identification Document AuthorsLanguage identification Document LanguagesText Classification Document Topics


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.


Author Identification

• Jane Austen (1775-1817), Pride and Prejudice

• Charles Dickens (1812-70), Bleak House



• 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


Author IdentificationFunction Words




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 Rights, UN, in 363 languages

http://www.unhchr.ch/udhr/navigate/alpha.htm


Language Identification

• égaux - French• eguali - Italian• iguales - Spanish

• edistämään - Finnish

• għ - Maltese


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

http://news.google.com/




Reuters Newswire Corpus


Reuters Sample<REUTERS TOPICS="YES" LEWISSPLIT="TRAIN" CGISPLIT="TRAINING-

SET" OLDID="12981" NEWID="798"><DATE> 2-MAR-1987 16:51:43.42</DATE><TOPICS><D>livestock</D><D>hog</D></TOPICS><TITLE>AMERICAN PORK CONGRESS KICKS OFF TOMORROW</TITLE><DATELINE> CHICAGO, March 2 - </DATELINE><BODY>The American

Pork Congress kicks offtomorrow, 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

</BODY></TEXT></REUTERS>


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


Classification

Class1

Class2


Classification

Class1

Class2


Classification

Class1

Class2


Classification

Class1

Class2


Clustering


Clustering


Clustering


Clustering


Clustering


Categories (Labels, Classes)

• Labeling data• 2 problems: • Decide the possible classes (which ones,

how many)– Domain and application dependent– http://news.google.com

• Label text– Difficult, time consuming, inconsistency

between annotators




Reuters• <REUTERS TOPICS="YES" LEWISSPLIT="TRAIN" CGISPLIT="TRAINING-SET"

OLDID="12981" NEWID="798">• <DATE> 2-MAR-1987 16:51:43.42</DATE>• <TOPICS><D>livestock</D><D>hog</D></TOPICS>• <TITLE>AMERICAN PORK CONGRESS KICKS OFF TOMORROW</TITLE>• <DATELINE> CHICAGO, March 2 - </DATELINE><BODY>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

• </BODY></TEXT></REUTERS>

Why not topic = policy ?


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


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


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


LabeledText class in NLTK

• LabeledText class

• >>> 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”

http://nltk.sourceforge.net/api/public/nltk.classifier.LabeledText-class.html


NLTK Classifier Interface• classify determines which label is most appropriate for a given text

token, and returns a labeled text token with that label. • labels returns the list of category labels that are used by the

classifier. • >>> 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",…)

http://nltk.sourceforge.net/api/public/nltk.classifier.ClassifierI-class.html#classify

http://nltk.sourceforge.net/api/public/nltk.classifier.ClassifierI-class.html#labels


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?


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?


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, mainland China)


Features

• Boolean (or Binary) Features• Features that generate boolean (binary) values. • Boolean features are the simplest and the most

common type of feature.

– f1(text) = 1 if text contain “Kerry”

0 otherwise– f2(text) = 1 if text contain PERSON

0 otherwise


Features

• Integer Features• Features that generate integer values. • Integer features can be used to give classifiers

access to more precise information about the text.

– f1(text) = Number of times text contains “Kerry”

– f2(text) = Number of times text contains PERSON


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.

– >>> 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

http://nltk.sourceforge.net/api/public/nltk.classifier.feature.FeatureDetectorI-class.html#detect


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)


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!


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


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


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)


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?


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)


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 ci – 2) all other documents

• the characteristic would be:– “document contains term t”


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 = auto 2 500

Class auto 3 9500


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: Pr(j,a) = Pr(j) Pr(a)– So, there would be: N Pr(j,a), i.e. N Pr(j) Pr(a)– Pr(j) = (2+3)/N; Pr(a) = (2+500)/N; N=2+3+500+9500– Which is: N(5/N)(502/N)=2510/N=2510/10005 0.25


Class = auto 2 500

Class auto 3 9500


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: Pr(j,a) = Pr(j) Pr(a)– So, there would be: N Pr(j,a), i.e. N Pr(j) Pr(a)– Pr(j) = (2+3)/N; Pr(a) = (2+500)/N; N=2+3+500+9500– Which is: N(5/N)(502/N)=2510/N=2510/10005 0.25


Class = auto 2 (0.25) 500 (502)

Class auto 3 (4.75) 9500 (9498)

expected: fe

observed: fo


2 statistic (CHI)2 is interested in (fo – fe)2/fe summed over all table entries:

The null hypothesis is rejected with confidence .999,

since 12.9 > 10.83 (the value for .999 confidence).

)001.(9.129498/)94989500(502/)502500(

75.4/)75.43(25./)25.2(/)(),(22

2222

p

EEOaj


Class = auto 2 (0.25) 500 (502)

Class auto 3 (4.75) 9500 (9498)

expected: fe

observed: fo


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:


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


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


TF.IDF Term Weighting

• TF: term frequency– definition: TF = tij

• 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/ni)

• ni : number of documents containing term i • N : total number of documents

– purpose: makes rare words across documents more important

• TF.IDF– definition: tij log(N/ni)


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)


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”?)


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 IDF4. Dimensionality reduction. (occasionally)


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


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


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


Linear vs. Non-Linear

• Linearly separable data: if all the data points can be correctly classified by a linear (hyperplanar) decision boundary



Linear Decision boundary



Class1Class2Non-Linearly Separable



Non Linear Classifier Class1Class2



• 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)



• 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)


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



• Data: {(xi,yi)}i=1...n

– x in Rd (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



• Find a good hyperplane

(w,b) in Rd+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)


Perceptron Algorithm• Initialize: w1 = 0• Updating rule For each data point x

– If class(x) != decision(x,w)

– then wk+1 wk + yixi

k k + 1 – else

wk+1 wk

• Function decision(x, w)– If wx + b > 0 return +1– Else return -1

wk

0

+1

-1wk x + b = 0

wk+1

Wk+1 x + b = 0


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”


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)


Winnow Algorithm• Initialize: w1 = 0• Updating rule For each data point x

– If class(x) != decision(x,w)– then

wk+1 wk + yixi Perceptron

wk+1 wk *exp(yixi) Winnow

k k + 1 – else

wk+1 wk

• Function decision(x, w)– If wx + b > 0 return +1– Else return -1

wk

0

+1

-1wk x + b= 0

wk+1

Wk+1 x + b = 0


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


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


Support Vector Machine (SVM)

• Large Margin Classifier

• Linearly separable case

• Goal: find the hyperplane that maximizes the margin

wT x + b = 0

M wTxa + b = 1

wTxb + b = -1

Support vectors


Support Vector Machine (SVM)

• Text classification

• Hand-writing recognition

• Computational biology (e.g., micro-array data)

• Face detection

• Face expression recognition

• Time series prediction


Classification II

• Non-linear algorithms

• Kernel methods

• Multi-class classification

• Decision trees

• Naïve Bayes

• Last topic for today: k Nearest Neighbour


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


1 Nearest Neighbour


1 Nearest Neighbour


3 Nearest Neighbour


3 Nearest Neighbour

Assign the category of the majority of the neighbors

But this is closer..We can weight neighbors according to their similarity


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

Documents

October 2005CSA3180: Text Processing III1 CSA3180: Natural Language Processing Text Processing 3 – Double Lecture Discovering Word Associations Text Classification