ai seminar

August 29, 2001 Melanie Martin - AI Seminar 1

AI Seminar

Our web page is at:

www.cs.nmsu.edu/~gradrep

Under “Events” in left frame


Identifying Ideological Point of View

Melanie Martin

August 29, 2001


Outline of this presentation What is AI??? Introduction and Motivation The proposed system Ideology Discourse Statistical NLP and Machine Learning Internet Conclusion


What is AI???

“The practice of designing systems that possess and acquire knowledge and reason with knowledge.” (Tanimoto 1987)

“The design and study of computer programs that behave intelligently.” (Dean, Allen, Aloimonos 1995)

“The branch of computer science concerned with making computers behave like humans.” (Webopedia)


What is AI???

But then, what is intelligence???– “the capacity for learning, reasoning,

understanding, and similar forms of mental activity; aptitude in grasping truths, relationships, facts, meanings, etc.” (Webster’s Encyclopedic Unabridged Dictionary of the English Language 1996)


What is AI???

Agents Data Mining Expert Systems Games and Search

Knowledge Representation

Machine Learning Theory, Case-Based, Rule

Learning, ...

Natural Language Processing Planning

Robotics

Speech Theorem Proving

Vision & Pattern Recognition

Categories under AI on Cora

http://cora.whizbang.com/


What is AI???

Goals in AI– Engineering: Solve real-world problems.

Build systems that exhibit intelligent behavior.

– Scientific: Understand what kind of computational mechanisms and knowledge are needed for modeling intelligent behavior.


What is AI??? Do we really want to model humans?

– Seem like our best example, but….– Should we build airplanes with wings that

flap like birds? How do we know we did it?

– Turing test?• Focus on behavior instead of internal algorithm• Defines success in terms of human intelligence• Not well founded


What is AI???

A couple of recurring issues:– How important is cognitive modeling in our

systems?– How do we balance scientific and

engineering goals?– How do we evaluate our system?


What is AI???

So let’s get to the system we want to talk about today…..

This system will be in the area of Natural Language Processing aka Computational Linguistics


Introduction and Motivation

Your back hurts, so you go to the web to find out what you can do, but there is too much information!

You are still bothered by the Florida election results and want to read a few sample articles with differing points of view. How can you find them?



Suppose we could take information from web pages and Usenet newsgroups on a given topic and segment, classify or cluster it by ideological point of view…..

This talk is about what it might take to develop such a system.



Sounds like a cool toy, but would it make any research contribution?

Areas where it could contribute:– natural language understanding– information retrieval– information extraction– internet structure



But will it save the world?

Maybe not, but there is social value in analyzing ideological point of view– find implicit ideological content– better informed, more rational discussion of

important issues


The Proposed System

Let’s recall what we want to do:

Build a system that could take information from web pages and Usenet newsgroups on a given topic and segment, classify or cluster it by ideological point of view…..


The Proposed System

IdeologicalClassifier

TopicClassifier,

Filter

Set of documents

on topic

Internet:Web pages,

Usenet

Docs ontopic

classified by IPV

SearchEngine

User inputstopic


The Proposed System

Immediately some issues arise:– Can we come up with a definition of

ideological point of view that is computationally feasible?

– To what extent do we need to understand the text?

– Would modeling human text understanding help?


The Proposed System

More issues:– Can the structure of the internet help us?– What kind of knowledge is needed and can

it be learned?– How are we going to evaluate our system?


Ideology

Working definition from van Dijk: “Ideologies are the fundamental beliefs of a group and its members.”– No negative evaluation– Subjective, since beliefs are subjective– Discourse plays a key role in development

and promulgation of ideologies


Ideology What do we mean by groups?

– More than one person– Fewer than the entire society or culture– Some level of permanency or common

goals– Some membership criteria– Member identification with the group– Basis for self-definition and commonality– Structure, possibly informal


Ideology

General strategy of most ideological discourse (van Dijk’s Ideological Square):

– Emphasize positive things about Us– Emphasize negative things about Them– De-emphasize negative things about Us– De-emphasize positive things about Them

Polarization; Us versus Them


Ideology

How are these strategies instantiated in discourse?– What is there:

• argument structure• syntactic patterns• style and non-literal language• actor descriptions• thematic structure• topoi


Ideology

– What is not there• implication• presupposition• inference• goals and plans


Ideology

Disclaimers, selected examples:– Apparent Negation: I have nothing against X, but...– Apparent Concession: They may be very smart,

but...– Apparent Empathy: They may have had problems,

but...– Apparent Effort: We do everything we can, but...

Positive self-representation and face keeping


Ideology

Linguistics– van Dijk (1998)– Blommaert & Verschueren (1998)– Wang (1993)– Wortham & Locher (1996)


Ideology

The Systems– Ideology Machine -1965 to 1973 - Abelson et al.– Tale-Spin - 1976 - Meehan– Politics - 1979 - Carbonell– Pauline - 1987 - Hovy– Viewgen - 1991 - Ballim & Wilks– Tracking Point of View in Narrative - 1994 - Wiebe– Spin Doctor - 1994 - Sack– Terminal Time - 2000 - Mateas et al.


Ideology

Some issues– Evaluation!!!– Hard-coded knowledge– Domain dependence– Cognitive plausibility– More precise definitions


Ideology

What do we want to take with us?– van Dijk’s definitions augmented by Sack

and Wiebe– mine everything for clues to ideological

point of view


Discourse

Now that we have a working definition of ideology and some ideas about things that might be clues, the question becomes how to find them?

First we are going to look at theories of discourse structure that might be useful.


Discourse

Computational Linguistics – Hobbs (1979)– Mann & Thompson (RST) (1988)– Grosz & Sidner (G&S) (1986)– Morris & Hirst (Lexical chains) (1991)

Psycholinguistics– Kintsch (1994)


Discourse Issues

– do we need it at all?– implementation

• Hobbs, G&S, RST

– finite number of fixed primitives• Hobbs, RST

– world knowledge• Hobbs

– domain specific


Discourse

A reasonable first approach: Lexical Chains (Morris & Hirst)

Sequences of related words spanning a topical unit in the text– based on lexical cohesion– encapsulates context– helps identify key phrases


Discourse

Lexical chains could help us in:– topic segmentation– intentional structure– lexical features for a classifier


Discourse

Lexical chains are easy to implement, but are unlikely to be sufficient…

For the next approximation: RST– Marcu’s implementation incorporating G&S– Mostly used for summarization and

generation– Would help get at the argument structure

of the text


Discourse

Would most likely use RST to generate features for a classifier or as input to a pattern recognizer

Nuclei spans help pick out the more important segments of text

Produces a tree that gives the structure of the rhetorical structure of the text


Discourse

None of the discourse theories look like the are going to stand alone– may be able to give us structural, lexical

and other features – need to consider classification or clustering

based on these features– so we turn to….


Statistical NLP and ML

Two techniques we will consider– Latent Semantic Analysis– Probabilistic Classification



Issues– clustering versus classification

• categories may not be predefined• may want to take a variety of features into

account

– favor learning over hard-coding knowledge– supervised versus unsupervised

• cost of annotated training data



Latent Semantic Analysis– text represented as a matrix

• entries are weighted frequency of word in context

– semantic space obtained through SVD• words appearing in similar context have similar

feature vectors

– characterizes semantic content of words in context



Why LSA is a good choice here– semantics is key component of ideological

discourse– clustering without need for predefined

categories– already shown useful for:

• summarization (Ando 2000)• text segmentation (Choi 2001)• measuring text coherence (Foltz 1998)



But LSA doesn’t use all of the stuff we just spent all this time talking about…

What if it doesn’t work very well? Another option is a probabilistic

classifier– assigns most probable class to an object

bases on a probability model



Probability model– defines joint distribution of variables

• set of feature variables and a class variable

Wiebe and Bruce (1995) got around the issue of not knowing the classes in advance by breaking up the problem and using a series of classifiers



Maybe this will work after all and we can use some of the features we have been talking about

Deciding which features to use can be determined statistically with goodness of fit of graphical models



Both methods seem to have a lot of potential

LSA would be easier to implement – possibly a baseline for evaluation of

probabilistic classifiers Less linguistic knowledge gain likely

with LSA


Internet

We would like to mine the structure of the internet – see if there is a correspondence with

groups– improved IR by topic– figure out what search engine to use as a

base for our system


Internet

Structure papers– Kleinberg (1997)– Kleinberg et al. (1999)– Terveen et al. (1999)– Whittaker et al. (1998)


Internet

Issues– topic or query disambiguation– what is a minimal unit– how to use the structure of the web

• finding authorities• communities and subgraphs

– Evaluation!!!


Internet

Kleinberg (1997)– link based model– hub - links to many related authorities– authority– iterative weighting algorithm that

converges (rapidly in practice)– can disambiguate authorities by sense– can be used to trawl for cyber communities


Conclusion It seems that such a system can be built

– find a good search engine– use Kleinberg’s algorithm to improve

collection of documents retrieved– use LSA and/or a probabilistic classifier to

handle the ideological point of view– with a probabilistic classifier use features

discussed in the ideology and discourse sections


The End

Thanks for listening!

If you want to know more, my Comprehensive Exam paper is at:

www.CS.NMSU.Edu/~mmartin/courses/comps_all.html

ai seminar

Documents

ai seminarour web page

natural language unde

english language

modeling intelligent

rule learning

web pages

cognitive modeling

engineering goals