Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
73 views

Natural Language Processing For Prolog Programmers

The document reviews the book "Natural Language Processing for Prolog Programmers" by Michael A. Covington, describing it as a textbook that presents simple natural language processing techniques for experienced Prolog programmers through examples and exercises, however it is criticized for its shallow treatment of linguistic theories and focus on programming solutions over conceptual understanding.

Uploaded by

Eliakim Gama
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
73 views

Natural Language Processing For Prolog Programmers

The document reviews the book "Natural Language Processing for Prolog Programmers" by Michael A. Covington, describing it as a textbook that presents simple natural language processing techniques for experienced Prolog programmers through examples and exercises, however it is criticized for its shallow treatment of linguistic theories and focus on programming solutions over conceptual understanding.

Uploaded by

Eliakim Gama
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/2476111

Natural Language Processing for Prolog Programmers

Article · July 2002


Source: CiteSeer

CITATIONS READS

59 4,370

3 authors, including:

Ken Barker Stan Szpakowicz


IBM Research University of Ottawa
70 PUBLICATIONS 1,416 CITATIONS 190 PUBLICATIONS 5,652 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Stan Szpakowicz on 18 July 2014.

The user has requested enhancement of the downloaded file.


Book Reviews

Natural Language Processing for Prolog Programmers


Michael A. Covington
(University of Georgia)

Englewood Cliffs, NJ: Prentice-Hall,


1994, xvi+348 pp; hardbound, ISBN
0-13-629213-5, no price listed

Reviewed by
Ken Barkerand Stan Szpakowicz
University of Ottawa

The title says it all: it is truly a textbook for Prolog programmers, for better and for
worse. It presents simple NLP for experienced Prolog programmers.
The book is written in a casual (occasionally quite casual) style and is eminently
readable if you are reasonably comfortable with Prolog. The presentation has a satis-
fying flow. The book feels well planned, and is certainly well executed, considering its
goals stated in the preface. Editing is admirably meticulous, formatting nearly flaw-
less. The many examples in Prolog are clear and work well. Prolog code for examples
is available by ftp. References are abundant and useful.
The book has nine chapters and two appendices:

1. Natural language
2. Templates and keywords
3. Definite-clause grammars
4. English phrase structure
5. Unification-based grammar
6. Parsing algorithms
7. Semantics, logic, and model theory
8. Further topics in semantics
9. Morphology and the lexicon
A. Review of Prolog
B. String input and tokenization

Chapter I offers a definition of NL and NLP, and a review of the usual linguistic strata.
It also advocates Prolog as an NLP tool. Chapter 2 offers a pattern-matching approach
to NL interfaces to operating systems and databases. Chapter 3 introduces parse trees,
phrase-structure rules, and Prolog's DCG notation; the chapter also hints at parsing
and semantic analysis (treated separately in Chapters 6, 7, and 8). Chapter 4 devel-
ops a set of phrase-structure rules for the main English parts of speech; it also goes
beyond phrase-structure rules and gives a simple presentation of ID/LP rules and
transformational grammar. Chapter 5 introduces a unification-based grammar formal-
ism and an extension to Prolog for this. Chapter 6 describes and compares top-down,
bottom-up, left-corner, and chart parsing as well as Earley's algorithm. Chapter 7 of-
fers techniques for representing, in Prolog, knowledge from English sentences. Chap-

137
Computational Linguistics Volume 22, Number 1

ter 8 discusses language translation, word-sense disambiguation, and understanding


events. Chapter 9 presents some background on morphology as well as some models
and techniques for computational morphology. Appendix A is a thorough review of
Prolog (it accounts for 10% of the book!). Appendix B provides Prolog code for string
input and tokenization, used elsewhere in the text.
This is not the first book that deals with NLP and Prolog. Gazdar and Mellish
(1989) wrote a book primarily on NLP, which uses Prolog to illustrate linguistic con-
cepts. The programming language is entirely secondary--there are parallel versions
of the same textbook that use Lisp and Pop-11. Pereira and Shieber (1987) proposed
to do subtle, advanced NLP in Prolog. Although their linguistic decisions might not
appeal to everybody, they are never shallow. Pereira and Shieber's approach was the
kind of introduction to NLP and Prolog that we find appealing to people seriously
interested in computational linguistics, who need a software tool.
Covington's book can be perhaps characterized as a Prolog book that concentrates
on NL-related issues. Readers familiar with his co-authored Prolog Programming in
Depth (Covington, Nute, and Vellino 1988) will appreciate the similarities between its
Chapter 13 ("Natural language processing") and the book under review. The latter
may even be thought of as an extension of the former.
Covington accurately describes the present book (p. xv) as "first and foremost a
book of techniques . . . . It is not a comprehensive handbook of NLP." However, while it
indeed is not a comprehensive handbook, it does attempt to cover or at least mention
in passing an astounding number of NLP issues and problems. Unavoidably, many of
them get a rather skin-deep treatment. One problem with such an approach is that a
book of techniques without depth risks coming across as a cookbook. And although
cookbooks often have good recipes, they don't necessarily teach you much about good
cooking.
Here's an example of a technique overshadowing theory. In the book there is an
implementation of lambda calculus for compositional semantics. The implementation,
however, is merely a neat Prolog device that gets Prolog to unify the variable argu-
ments of structures with different functors. After the briefest explanation (section 2.3.6)
"lambda" becomes merely a name for a specific form of Prolog terms. Not a word on
why lambda calculus can represent certain linguistic phenomena, nor on what those
phenomena are.
The Prolog techniques for NLP are presented in a fairly consistent format:

• here is an NLP problem;


• here is the author's simple solution in Prolog;
• here are exercises for the reader to improve on the solution.

This is not, perhaps, a format that stimulates deeper understanding. Solutions thrust
upon the reader risk being memorized without due generalization.
The exercises are heavily biased towards Prolog. Too many of them are actually
Prolog problems with little to do with linguistics. Some of them ask the reader to
figure out why some piece of Prolog code works. This usually requires a good knowl-
edge of Prolog and maybe even Prolog implementations. Nothing to it, if you are a
Prolog programmer (fair enough; just recall the book's title) and you are willing to
buy off-the-shelf components and refrain from delving into the underlying theories
and algorithms.
Numerous exercises call for maintaining or enlarging the author's ready-made
solutions. The reader is not really expected to come up with her own ideas: there is

138
Book Reviews

no challenge to think linguistically, though there is challenge to think Prologically.


A fresh convert to NLP may find that the NL issues discussed in the book often
appear to be very easily solved. For example, section 8.2 introduces machine transla-
tion, produces a solution, and gives exercises, in the space of four pages. The treatment
looks so enticingly simple that the author himself found it necessary to have an ex-
tra page at the end (section 8.2.5) explaining "Why translation is hard". The same
goes, mutatis mutandis, for section 8.3 ("Word sense disambiguation') and section 8.4
("Understanding events"). It is small consolation that the author tells us (p. 233) this
chapter will be "a whirlwind tour." It is not. It may be a bird's-eye view, but the bird
is no falcon.
In a way, the very same concern holds for the whole book. The author says (p. xv):
"It is meant for computer science students and working programmers who know
Prolog reasonably well but have little or no background in linguistics." We are not
sure whether they will come away from this book with much more than they arrived
with. They may think they do, lured by the seeming facility of all those elegant, succinct
Prolog morsels. However, unless additional material is covered in a course that uses
this textbook, the reader will have only seen many short and sweet programs and
many learned technical terms.
The book should be of good use in a specialized undergraduate course, and could
also serve as a dependable technical supplement in a linguistically weightier graduate
course for students in computing. Apart from such students, the book may appeal to
a very limited audience: people who know Prolog well and may need some NLP for
their mainstream programming or applications. Linguists might be overwhelmed by
the technicality, and if they got past the technicality, the linguistic content remaining
after stripping away the programming layer might not turn out to be too edifying
for specialists. Perhaps they would appreciate Chapter 9 with its clear and appealing
presentation. On the other hand, they might miss a serious mention of some real NLP
systems.
Covington asks us (p. xv) to "judge this book by what it contains, not by what it
leaves out." What it leaves out is deep linguistic motivations. What it contains is some
clever and useful Prolog solutions to simplified NLP problems. So if that's what you
need, this book is for you.

References Introduction to Computational Linguistics.


Covington, Michael; Nute, Donald; and Wokingham: Addison-Wesley.
Vellino, Andr4. (1988). Prolog Programming Pereira, Fernando C. N. and Shieber, Stuart,
in Depth. Glenview, IL: Scott, Foresman. M. (1987). Prolog and Natural-language
Gazdar, Gerald and Mellish, C. S. (1989). Analysis. Stanford: Center for the Study of
Natural Language Processing in Prolog: An Language and Information.

Ken Barker is a doctoral student in computational linguistics at the University of Ottawa. His the-
sis deals with building semantic networks from parse trees in Prolog in the absence of precoded
semantic knowledge. He has published one journal paper and three conference papers. Stan
Szpakowicz is a professor of Computer Science at the University of Ottawa. His research inter-
ests include text analysis for knowledge acquisition. He has published about 80 refereed papers
and four books. The authors' address is: Department of Computer Science, University of Ottawa,
Ottawa, Ontario, Canada KIN 6N5; e-mail: kbarker@csi.uottawa.ca and szpak@csi.uottawa.ca.

139
View publication stats

You might also like