438/538 Computational Linguistics Sandiway Fong Lecture 1: 8/22.

438/538Computational Linguistics

Sandiway Fong

Lecture 1: 8/22

Part 1

• Administrivia

Administrivia

• Where– BIO W 212

• When– TR 12:30–1:45PM

• No Class– Thursday September 14th– Thursday September 28th– Thursday November 23rd (Thanksgiving)

• Office Hours– catch me after class, or– by appointment– Location: Douglass 311

Administrivia

• Map

– Office (Douglass)

– Classroom (FCS)

Administrivia

• Email– [email protected]

• Homepage– http://dingo.sbs.arizona.edu/~sandiway

• Lecture slides– available on homepage after each class– in both PowerPoint (.ppt) and Adobe PDF formats

• animation: in powerpoint

– last year’s slides are available • (new material for this year will be rotated in)

Administrivia

Reference Textbook• Speech and Language Processing,

Jurafsky & Martin, Prentice-Hall 2000– 21 chapters (900 pages)– Concepts, algorithms, heuristics– Sound/speech side

• N. Warner Speech Tech LING 578 (this semester)

• Y. Lin Statistical NLP LING 539 (Spring 2007)

– Intersection with research areas• Parsing and Linguistic Theory

(Sentence Processing) • Computational Morphology• Machine Translation, WordNet

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.

Administrivia

• Course Objectives– Theoretical

• Introduction to a broad selection of natural language processing techniques

• Survey course• Relevance to linguistic theory

– Practical• Acquire some expertise

– Parsing algorithms– Write grammars and machines– Build a toy machine translation system

Administrivia

• Laboratory Exercises– To run tools and write grammars– you need access to computational facilities

• use your PC (Windows, Linux) or Mac

– Homework exercises

Administrivia

• Homeworks and Grading– 6~8 homeworks

• no final or mid-terms• mix of theoretical and practical exercises• there will be mandatory and extra credit questions

– extra credit questions matter: » make up points lost on other questions in the

homework» may bump you up a grade at the end of the semester

in borderline situations

• some simple programming is involved (no prerequisite)• use of a spreadsheet (Excel) for numerical calculation

Administrivia

• Homeworks and Grading– Homeworks will be presented/explained in class

• (good chance to ask questions)

– Please attempt homeworks early• (then you can ask questions before the deadline)

– Unless otherwise specified, you have one week to do the homework

• (midnight deadline)• (email submission to me)• e.g. homework comes out on Thursday, it is due in my mailbox

by next Thursday midnight

– Look for acknowledgement email from me

Administrivia

• Homework Ethics– you may discuss homework with your classmates– however, you must do the work and write them up independently– sources must be acknowledged,

• e.g. if you borrow program code off the internet• discovered cheaters will be sanctioned

• Late Policy– all homework is mandatory

• you can’t get an A skipping a homework• some homeworks may depend on earlier homeworks

– deductions if late– if you know you are going to be late, notify me ahead of time

• e.g. upcoming emergencies

Administrivia

• 438 vs. 538538

=

438

+

1 classroom presentation of a selected chapter

+438 extra credit homework questions are obligatory

Administrivia

• There is a laptop being passed around

• Fill out spreadsheet entry– Name– Email– Year/Major– 438 or 538– Relevant background

Administrivia

• Class demographics (8/20 classlist)438/538

LING

COSC

EPH

MATHNDSNMS

MGT

PSYC

SLA

ESL

APPL

Part 2

• Introduction

Human Language Technology (HLT)

• ... is everywhere

• information is organized and accessed using language

Human Language Technology (HLT)

Beginnings• c. 1950 (just after WWII)

– Electronic computers invented for• numerical analysis• code breaking

Killer AppsKiller Apps: : – Language comprehension tasks and Machine Translation (MT)Language comprehension tasks and Machine Translation (MT)

Reference– Readings in Machine Translation– Eds. Nirenburg, S. et al. MIT Press 2003. – (Part 1: Historical Perspective)

Human Language Technology (HLT)

• Cryptoanalysis Basis– early optimism

[Translation. Weaver, W.]• Citing Shannon’s work, he asks: • “If we have useful methods for solving almost any cryptographic

problem, may it not be that with proper interpretation we already have useful methods for translation?”

Human Language Technology (HLT)

• Popular in the early days and has undergone a modern revival

The Present Status of Automatic Translation of Languages (Bar-Hillel, 1951)

– “I believe this overestimation is a remnant of the time, seven or eight years ago, when many people thought that the statistical theory of communication would solve many, if not all, of the problems of communication”

– Much valuable time spent on gathering statistics• perhaps no longer a bottleneck

Human Language Technology (HLT)

• uneasy relationship between linguistics and statistical analysis

Statistical Methods and Linguistics (Abney, 1996)

– Chomsky vs. Shannon

• Statistics and low (zero) frequency items– Smoothing

• No relation between order of approximation and grammaticality

• Parameter estimation problem is intractable (for humans)– IBM (17 million parameters)

Human Language Technology (HLT)

• recent exciting developments in HLT– precipitated by progress in

• computers: stochastic machine learning methods• storage: large amounts of training data

– recent improvements in stochastic models from incorporating linguistic knowledge

– (Hovy, MT Summit 2003)

Human Language Technology (HLT)

• Killer Application?

Natural Language Processing (NLP)Computational Linguistics

• Question:– How to process natural languages on a computer

• Intersects with:– Computer science (CS)– Mathematics/Statistics – Artificial intelligence (AI)– Linguistic Theory– Psychology: Psycholinguistics

• e.g. the human sentence processor

Natural Language Properties

which properties are going to be difficult for computers to deal with?

• Grammar (Rules for putting words together into sentences)– How many rules are there?

• 100, 1000, 10000, more …

– Portions learnt or innate– Do we have all the rules written down somewhere?

• Lexicon (Dictionary)– How many words do we need to know?

• 1000, 10000, 100000 …

Computers vs. Humans

• Knowledge of language– Computers are way

faster than humans• They kill us at arithmetic

and chess

– But human beings are so good at language, we often take our ability for granted

• Processed without conscious thought

• Exhibit complex behavior

QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.

IBM’s Deep Blue

Examples

• Innate Knowledge?– Which report did you file without reading?– (Parasitic gap sentence)– file(x,y)– read(u,v)

x = youy = reportu = x = youv = y = reportand there are no other possible interpretations

*the report was filed without reading

Examples

• Changes in interpretation• John is too stubborn to talk to• John is too stubborn to talk to Bill

talk_to(x,y)

(1) x = arbitrary person, y = John

(2) x = John, y = Bill

Examples

• Ambiguity– Where can I see the bus stop?

– stop: verb or part of the noun-noun compound bus stop– Context (Discourse or situation)

– Where can I see [the [NN bus stop]]?– Where can I see [[the bus] [V stop]]?

Examples

• Ungrammaticality– *Which book did you file the report without

reading?

– * = ungrammatical• relative

– ungrammatical vs. incomprehensible

Example

• The human parser has quirks• Ian told the man that he hired a story• Ian told the man that he hired a secretary

• Garden-pathing• Temporary ambiguity• tell: multiple syntactic frames for the verb

• Ian told [the man that he hired] [a story]• Ian told [the man] [that he hired a secretary]

Examples

• More subtle differences• The reporter who the senator attacked admitted the error• The reporter who attacked the senator admitted the error

– Processing time differences• Subject vs. object relative clauses

– Q: Do we want to mimic the human parser completely?

Frequently Asked Questions from the Linguistic Society of America (LSA)

• http://www.lsadc.org/info/ling-faqs.cfm

LSA (Linguistic Society of America) pamphlet

• by Ray Jackendoff

• A Linguist’s Perspective on What’s Hard for Computers to Do …

– is he right?

QuickTime™ and aTIFF (Uncompressed) decompressorare needed to see this picture.

If computers are so smart, why can't they use simple English?

• Consider, for instance, the four letters read; they can be pronounced as either reed or red. How does the machine know in each case which is the correct pronunciation? Suppose it comes across the following sentences:

• (l) The girls will read the paper. (reed) • (2) The girls have read the paper. (red) • We might program the machine to pronounce read as reed if it

comes right after will, and red if it comes right after have. But then sentences (3) through (5) would cause trouble.

• (3) Will the girls read the paper? (reed) • (4) Have any men of good will read the paper? (red) • (5) Have the executors of the will read the paper? (red) • How can we program the machine to make this come out

right?

If computers are so smart, why can't they use simple English?

• (6) Have the girls who will be on vacation next week read the paper yet? (red)

• (7) Please have the girls read the paper. (reed)• (8) Have the girls read the paper?(red)• Sentence (6) contains both have and will before read, and both

of them are auxiliary verbs. But will modifies be, and have modifies read. In order to match up the verbs with their auxiliaries, the machine needs to know that the girls who will be on vacation next week is a separate phrase inside the sentence.

• In sentence (7), have is not an auxiliary verb at all, but a main verb that means something like 'cause' or 'bring about'. To get the pronunciation right, the machine would have to be able to recognize the difference between a command like (7) and the very similar question in (8), which requires the pronunciation red.

Next time …

• We will begin by introducing you to a programming language you will become familiar with– Two introductory lectures– Name: PROLOG (Programming in Logic)– Variant: SWI-PROLOG (free software from University of

Amsterdam)• Download: http://www.swi-prolog.org/• Install it on your PC or Mac

– Based on mathematical logic• logic and inference are useful tools

– Contains built-in grammar rules• programming language was originally designed for NLP

Prolog Resources

• Some background in programming?

• Useful Online Tutorials– An introduction to Prolog

• (Michel Loiseleur & Nicolas Vigier)• http://invaders.mars-attacks.org/~boklm

/prolog/

– Learn Prolog Now! • (Patrick Blackburn, Johan Bos &

Kristina Striegnitz)• http://www.coli.uni-saarland.de/~kris/le

arn-prolog-now/lpnpage.php?pageid=online

Prolog Resources

• No background at all?

• Audit– LING 388 Computers and Language

• (also taught by me)• first couple of weeks• introduces Prolog at a more gentle pace • uses lab classes for practice

• Lectures TR 3:30–4:45pm– Harvill 313

• Hands-on Lab Class: this Thursday– Social Sciences 224