learning from multi-topic web documents for contextual advertisement kdd 2008

Learning from Multi-topic Web Documents for Contextual

Advertisement

KDD 2008

112/04/20 2

Outline

1. Introduction

2. Sensitive Content Detection

3. Sentiment Classification and Detection & Opinion Mining

4. Experiments

5. Conclusion

112/04/20 3

1. Introduction (1/4)

• Contextual advertisement – A popular advertising paradigm where web page owners

allow ad platforms to place ads on their pages that match the content of their sites

– Problems:• The huge variety of content that can appear on a single web page

– e.g. news sites, blogs, etc

• Advertisers do not want to show their ads on pages with content like violence, pornography etc. (sensitive content)

• They may also not wish to advertise on pages which contain negative opinion about their products

112/04/20 4


• Objective– Not only to tell if a document has some targeted

content in it, but also to label the parts of the document where the content is present

• Sub-document classification– Classifier train on entire pages using page-level

labels and test on individual blocks

112/04/20 5


• Challenges:– Pages can contain unwanted parts

• e.g., navigation panes, text advertisements, etc

– Pages may also contain information on multiple topics

– To collect large amounts of broad coverage single-topic training data, pre-clean and hand-label the blocks are difficult and expensive

112/04/20 6


• This paper using multiple-instance learning (MIL) techniques – MILBoost to improve the performance of traditional methods (Naive-Bayes and Decision tree)

• To train sub-document classifiers using only page level labels

• The problems of sensitive content detection and opinion/sentiment classification for advertising can be considered as 2-class and multi-class classifying

• In sentiment detection, a Naive-Bayes based MILBoost detector performs as well as the best block detector trained with block-level labels

112/04/20 7

Outline

1. Introduction



4. Experiments

5. Conclusion

112/04/20 8

2. Sensitive Content Detection (1/3)

• Sensitive content categories – e.g., crime, war, disasters, terrorism, pornography, etc

• The various sensitive categories are grouped as one class labeled as “sensitive”

• As long as a web page contains such content blocks, it will be marked as sensitive and the ad display will be turned off

• The available training web pages are labeled at the page-level– The labels only tell whether a page contains sensitive

content somewhere in it or not

112/04/20 9


112/04/20 10


• If using the entire page, traditional classification methods run the risk of learning everything on the page as “sensitive”

• To avoid this problem, a classifier that can accurately identify the parts of the page that contain the targeted content is needed

• Better still is a classifier that can integrate the two tasks of locating and learning – Multiple Instance Learning framework

112/04/20 11

2.1. Multiple Instance Learning Boosting (1/8)

• Multiple Instance Learning (MIL) is a variation of supervised learning where labels of training data are incomplete

• Traditional methods where the label of each individual training instance is known

• In MIL the labels are known only for groups of instances – Bag: a web page

– Instance: a block of text

112/04/20 12


• 2-class classification (sensitive or non-sensitive)– A bag is labeled positive if at least one instance in that bag

is positive

– A bag is labeled negative if all the instances in it are negative

• The goal of MIL algorithm is to produce a content detector at the sub-document (block) level without having the block labels in the training data

• This can save significant amount of money and effort by avoiding labeling work at the block level

112/04/20 13


• Why MILBoost:– The state of the art traditional algorithms use boosting

– We needed a framework to accurately measure the added effectiveness of the MIL framework

– MILBoost has been successfully applied to a similar problem

• Training a face detector to detect multiple faces in pictures when only picture level labels are available

112/04/20 14


Positive

Positive NegativePositive

NegativePositive

NegativeNegative

?

? ?

Positive

Positive

Positive

?

? ?

Training initial classifier

Positive

NegativeNegative

Negative

Negative

112/04/20 15


• For each instance of bag , the probability of an instance is positive is given by

where

is the weighted sum of the output of each classifier in ensemble with t steps

• is the output score of the instance generated by the classifier of ensemble

ijtt

tijij xCxCy )(

ijx ijxiB

ijij y

P

exp1

1

ijxtht)( ijt xC

112/04/20 16


• The probability that the bag is positive is a “noisy OR”

• Under this model the likelihood assigned to a set of training bags is

where is the label of bag i

ij

ijii PPP 111

i

li

li

ii PPCL )1()1()(

1,0il

112/04/20 17


• Following the AnyBoost approach, the weight on an instance is given as

• Each round of boosting is a search for a classifier which maximum

where is the score assigned to the instance of bag i by the weak classifier (for a binary classifier )

• The parameter is determined using a line search to

maximum

ijx

iji

ii

ijij P

P

Pl

y

CLw

)(log

ij

ijij wxc )(

)( ijxc ijx 1,1)( ijxc

1t)(log ttCCL

1tC

112/04/20 18


112/04/20 19

112/04/20 20

Outline

1. Introduction



4. Experiments

5. Conclusion and Future Work

112/04/20 21


• Sentiment/opinion mining from review pages or blogs – A page may contain one or more topics

– It is common to label reviews as “positive” or “negative”

– Reviews are often not as polar or one sided as the label indicates

– Blog review sites or discussion forums usually feature many people expressing varied opinions about the same product

– These “mixed” opinions may act as noise during the training of traditional classification methods

112/04/20 22

3.1 Multi-target MILBoost Algorithm (1/6)

• To apply MILBoost to the multi-topic detection task, it needs to be extended to a multi-class scenario

• The “positive” and “negative” opinions can be treated as the target classes and the “neutral” class as the null class

• A bag is labeled as belonging to class k if it contains at least one instance of class k

• A bag can be multi-labeled since it may contain instances from more than two different target classes

• To deal with multi-labels– Creating duplicates of a bag with multiple labels – Assigning a different label to each duplicate

112/04/20 23


• Suppose we have 1 . . .K target classes and class 0 is the null class

• For each instance of bag , the probability that belongs to class k (k {1, 2, . . . ,K}) is given by a softMax function,

where

is the weighted sum of the output of each classifier in the ensemble with t steps

• is the output score for class k from instance generated by the classifier of the ensemble

ijx

ijxiB

ijx tht)( ijkt xC

t

ijkttijk xCY )(

112/04/20 24


• The probability that a page has label k is the probability that at least one of its content block has label k

• Assuming that the blocks are independent of each other, the probability that a bag belongs to any target class k (k > 0) is

• The probability that a page is neutral (or belongs to the null class 0) is the same as the probability that all the blocks in the page are neutral

(“noisy OR” model)

112/04/20 25


• The log likelihood of all the training data can be given as

• The weight on each instance for next round of training is given as

• For the null class,

112/04/20 26


• Combining weak classifiers– Once the (t + 1)th classifier is trained, the weight on

the classifier can be obtained by a line search to maximize the log likelihood function

• Choice of classifier– In experiments, Naive Bayes and decision trees are

used

1t

tC

1t

112/04/20 27


• Testing– The new page is divided into blocks and the block level

probabilities are computed using the classifier

– The page level probabilities are obtained by combining the block level probabilities using noisy-OR

– The block and page level labels are calculated using thresholds on the probabilities

112/04/20 28

Outline

1. Introduction



4. Experiments

5. Conclusion

112/04/20 29

4.1 Sensitive Content Detection (1/5)

• The data set contains 2,000 web pages which are labeled at the page level by human annotators

• The label for each web page is binary, either sensitive or non-sensitive

• There is no labeling done at the text block level

• The evaluation has to be done at the web page level

• Two popular base classifiers were used to build the MILBoost ensemble, decision trees and Naive Bayes

• Both the MILBoost and the non-MILboost versions were run through 30 boosting iterations which end up with an ensemble of 30 classifiers

• Area Under the ROC Curve (AUC) was used to evaluate the effectiveness of the various detectors

112/04/20 30


Significantly outperforms both base classifiers and traditional boosted versionsMILBoost further improves this performance by another 8.2%The MILBoost version achieved almost the same performance as the

boosted page-classifier

112/04/20 31


Althoug the AUC is about the same, the MILBoosted system is almost consistently better than the boosted page-classifier at the early part, where usually the operation point exists. This “early lift” brings practical advantage to the MILBoosted system.

112/04/20 32


• Naive Bayes vs Decision Trees– Naive Bayes performed much better than decision trees

in this task

– The decision tree ensemble uses only about 700 keywords while NB theoretically uses the whole vocabulary, which is about 20,000

– The bigger feature set enables NB to generalize better at the testing stage

112/04/20 33


• A Sensitive Content Detection Demo

112/04/20 34

4.2.1 Sentence Level Sentiment Detection (1/2)

• The subjectivity dataset from the Cornell movie review data repository is used as the data set

• 10000 “objective” and “subjective” sentences are labeled

• These sentences were extracted from 3000 reviews, which are labeled at the review-level as well

• A review is a “page” and a sentence is a “block”

• The MILBoost detector is trained with the review data only using page-level labels, and then evaluated at the sentence-level with sentence level labels

• Traditional page-level classifiers using boosted NB and decision trees are also built as benchmark algorithms for comparison

• A page-level classifier using support vector machines (SVM) is also trained to compare the performance

112/04/20 35

4.2.1 Sentence Level Sentiment Detection (2/2)

Train the classifiers with the reviews using page-level labelsTrain the classifiers with the reviews using sentence-level labelsThe highest performance in all algorithms using page-level labels, and it is comparable with the best sentence detector trained with sentence-level labels

MILBoost improves the performance by about 10% over boosted decision treesThe SVM did not do as well as the NB classifiers for sentence classification trained either with the page- or with the sentence-level label.

112/04/20 36

4.2.2 Multi-class Sentiment Detection (1/3)

• Sentiment detection is naturally a three-class problem with “positive”, “negative” and “neutral” as class labels

• The “positive” and “negative” classes are the target classes and the “neutral” class is the null class in the MILBoost setup

• In these tasks, only built a multi-class MIL system based on Naives Bayes

• The evaluation can only be done at the page-level

112/04/20 37


MILBoost based system improves upop the boosted Naive Bayes classifierThe performance using SVM is comparable to the MILBoost system

112/04/20 38


112/04/20 39

4.2.3 When does MIL improve on traditional methods? An Analysis Experiment (1/3)

• This paper hypothesized before that multiple-instance learning should improve learning of traditional techniques when the amount of mixed content is high

• The experiments were run on a car review dataset which contained 113,000 user reviews from MSN Autos

• The objective of the learning task to identify negative opinions in review texts

• These experiments want to show that as the amount of mixed content increases, MIL based approach can help traditional techniques improve

112/04/20 40


• This data set had an overall review rating score from 0-10

• Assume that if the rating score is 6 or below, there will be some negative opinions in the review text

• Further split the negative reviews into two subsets, one with rating scores from 0 to 3 (data 0-3)and the other with ratings from 4 to 6 (data 4-6)

• Presumably, the percentage of negative sentences in “data 0-3” will be much higher than that in “data 4-6”

• If hypothesis hold right, MIL based techniques should give a bigger boost in “data 4-6”

112/04/20 41


The MILBoost based system did not improve much over the regular boosted system

Statistically significant improvement over traditional classifiers

With good quality training data, MILBoost does not give much advantage over traditional methods; however, if the training data has a high ratio of mixed content, then MILBoost does provide significant advantages

There are three times as many pages in “data 4-6” as in “data 0-3” and the entire class distribution is highly biased towards positive with positive to negative ratio of 5:1

112/04/20 42

Outline

1. Introduction



4. Experiments

5. Conclusion

112/04/20 43

5. Conclusion

• This paper explored sub-document classification for contextual advertisement applications where the desired content appears only in a small part of a multi-topic web document

• The sub-document classifiers are trained when only page level labels are available

• It showed that the MILBoost system improve the performance of the traditional classifiers in such tasks, especially when the percentage of mixed content is high

• These systems provide good quality block level labels for free, leading to significant savings in time and cost on human labeling at the block level

112/04/20 44

END

112/04/20 45

AdaBoost

112/04/20 46

Multi-labelled Document

learning from multi-topic web documents for contextual advertisement kdd 2008

Documents

content blocks

sensitive content categories

targeted content

multiple instance learning

pagelevel labels

huge variety of content

web page owners

pagelevelthe labels