I am writing to learn about some very broad fields, and I would like
someone to recommend key authors for me read in the areas of
Linguisitics, Psychology, and Computer Science.  Every time I try to
find what I'm looking for in these areas, I get lost down some
interesting but not quite relevant path.  So I'd like to tell you what
I want to learn and do, and maybe someone could tell me the quickest
and best things to read and study to get there.

I am thinking about going back to school to study for a Ph.D. in
Psychology.  I have a Master's in Clinical Psychology that was all
about clincal practice.  I learend a lot about how to treat patients,
but now I'm interested in doing a Ph.D. about something more academic.  I
have a background as a software engineer and as writer that I would to
use in my research.

In particular, I would like to know more about how linguists, and
particularly people in the field of Computational Linguistics, analyze
language.  I'd like to dig around and learn about how to analyze
transcripts of conversations to understand what the transcripts can
tell about the metaphors, meanings, and psychological processes of the
people who are talking.

I'd also like to learn about the methods someone could use to write
software to analyze transcripts.  My idea is that there is plenty of
research data out there from transcripts of psychotherapy sessions
that is correlated to outcomes.  I'd like to figure out how to write a
program that could linguistically analyze the transcripts to find
patterns that relate to outcomes.  I figure that I could probably get
a Ph.D. out of this, but I really have no idea where to start learning
the things I need to know.

I really have no idea where to begin.  I did study Chomsky and
anthropological linguistics years ago, but that seemed more related to
an innate structure of language, and not about what happens in
conversations.  I'd like someone to give me a list of authors, books,
and articles to get started.  I realize my question is huge, so I'd
like that list to be as narrow as possible.  I'm giving myself a year
or two before I apply to schools to try and figure out what I really
want to study.  I'll want to figure out as soon as possible if this
makes any sense as a research project.

I have a few basic assumptions that I need to confirm or deny:

1. I'm assuming that there are linguistic rules that exist that allow
you to discover broader conceptual or psychological processes in
conversation, and that somebody has written about them.

2. I am assuming that one could write software to analyze
conversational data based on these rules.  I'm guessing that someone
has written quite a bit about how to do this.

I know so little about these fields, that I suspect someone may have
already done work on this.  If so, please let me know who they are.

I guess that's about it.  Thanks!

Dear nefariouslacker2,


Both of your assumptions are correct. Within the broader discipline of
natural-language processing (NLP), there is a field called discourse
analysis, the premise of which is that the narrative meaning of a
natural-language text, whether written or spoken, can be extracted
by programmatic means. The goal of computational linguists who study
discourse analysis is to design and implement systems to do exactly that.

The discourse-analysis systems that have thus far been built are
experimental in nature, yielding results that are not nearly good
enough to be useful in a commercial setting. Far from discouraging
you, this state of affairs should tell you that the field is ripe for
exploration. The challenges are daunting and varied, and progress has
so far been limited, but there is strong demand and plenty of funding
for new research in discourse analysis.


To get an idea of the kind of algorithms that go into discourse analysis,
read this exploratory paper from 1994.

http://www-nlp.cs.umass.edu/ciir-pubs/aaai94_soderland.pdf
University of Massachusetts: Stephen Soderland and Wendy Lehnert:
Corpus-Driven Knowledge Acquisition for Discourse Analysis


A more recent paper that describes a real system used to solve a specific
problem in discourse analysis is the following.

http://www.isi.edu/natural-language/people/hovy/papers/04Coling-opinion-valences.pdf
Information Sciences Institute: Soo-Min Kim and Eduard Hovy: Determining
the Sentiment of Opinions


The above papers should give you the general flavor of discourse
analysis, although they are not concerned with the particular problem
of analyzing psychotherapeutic transcripts. The insights that you bring
to this original problem would make a splendid subject for a doctoral
dissertation. One good place to pursue your studies, by the way, would
be the Information Sciences Institute (ISI), which has the highest
concentration of NLP expertise in the United States and probably in the
world. The Director of the Natural Language Group at ISI is Ed Hovy,
who co-authored the second of the two papers above.

http://www.isi.edu/natural-language/people/hovy/bio.html
Information Sciences Institute: Eduard Hovy


Among Ed Hovy's many publications on computational linguistics is a
collection of academic essays, not written but co-edited by him, on
the topic of discourse analysis. It is available from several Amazon
Marketplace dealers.

http://www.amazon.com/exec/obidos/tg/detail/-/3540609482/qid=1131197364/sr=8-1/ref=sr_8_xs_ap_i1_xgl14/103-8257266-7362234?v=glance&s=books&n=507846
Amazon: Computational and Conversational Discourse : Burning Issues -
An Interdisciplinary Account


The other book I can recommend if you wish to read up on discourse
analysis is by Daniel Marcu, a colleague of Ed Hovy's. This book is
based on Marcu's doctoral research, which dealt with the comparative
merits of rule-based and statistical information extraction procedures
carried out on natural-language text.

http://www.amazon.com/exec/obidos/ASIN/0262133725/o/qid%3D974663607/sr%3D2-1/103-8257266-7362234
Amazon: The Theory and Practice of Discourse Parsing and Summarization


Most recent advances in computational linguistics have been made through
the statistical approach, and even rule-based systems today have some
statistical components. If you wish to start learning the mechanics
of NLP, therefore, you will want to invest in the following textbook,
which is a standard primer for students at both the undergraduate and
graduate levels.

http://www.amazon.com/exec/obidos/tg/detail/-/0262133601/ref=pd_bxgy_img_2/103-8257266-7362234?v=glance&s=books
Amazon: Foundations of Statistical Natural Language Processing


If and when you decide that discourse analysis really is your cup of
tea, and if you can relocate to the vicinity of Los Angeles, you should
consider applying for the PhD program in computational linguistics at
ISI. The Institute is located in Marina Del Rey, but most students live
around the USC campus in downtown Los Angeles.

http://www.isi.edu/natural-language/involved.html
Information Sciences Institute: Natural Language Group: Get Involved


It has been a pleasure to address this question on your behalf, and I
wish you all the best in your ambitious endeavor.

Regards,

leapinglizard

---

Request for Answer Clarification by nefariouslacker2-ga on 05 Nov 2005 12:23 PST

Dear Leapinglizard, 

Your answer has helped me a great deal with the linguistic and
computation part of my question, which was most of what I was asking. 
Thanks you.

However, I was also asking, if possible, to find a university setting
where people in Psychology are are actively applying this kind of
work, discourse analysis, specifically to psychotherapy and possibly
transcripts of sessions.  Forgive me if the faculty at ISI do have
background to support me in Linguistics, Computer Science, as well as
Clinical Psychology or Research Psychology.  From visiting their site,
it appears that they may not have much very many faculty to support
Clinical Psychology research.

I did a quick Web search on "discourse analysis", "psychotherapy", and
the word "computational", which came up with a few leads, including
some interesting work by the following folks:

- Bucci (Derner Institute of Advanced Psychological Studies, Adelphi
University, Glass Institute for Basic Psychoanalytic Research)
 
- B. Maskit (The University at Stony Brook)

- Roussos (Glass Institute for Basic Psychoanalytic Research)

You did a great job helping me to find some great information on the
linguistic and computation part of the question.  Sorry to be
hard-nosed about this, but for me, a $100 is a lot of money, and I´m
offering it cause this is a really hard question (at least for me). 
Please let me know what you find out about the psychotherapy
applications of computational discourse analysis.

Looking forward to your reply....

---

Clarification of Answer by leapinglizard-ga on 06 Nov 2005 23:04 PST

ISI does not itself have any clinical psychologists on staff. It is,
however, affiliated with the University of Southern California, which
has a large and reputable Department of Psychology. Students receive
their funding through ISI and do most of their research there, but all
coursework is done at USC. Because most work in computational
linguistics is by its very nature interdisciplinary, it would be
perfectly natural for an ISI student to take courses in psychology and
to procure clinical transcripts through the Department of Psychology.

University of Southern California: Department of Psychology
http://psychology.usc.edu/

Bear in mind that the problem of parsing psychotherapeutic
conversation with automatic methods is mostly a computational problem,
not a psychological one. I guarantee that no department of psychology
at any university in the world can teach you the computational
principles and algorithms you will need to tackle this problem. Your
best bet is to look for a solid program in computer science or NLP in
particular at a university that also has a serious department of
psychology. Take an advisor in computer science who specializes in
natural language processing, and form a liaison with a professor of
psychology to obtain transcripts for your research.

Stony Brook might be a good choice, by the way. It has a solid
reputation in both computer science and psychology. The HCI/NLP
computing lab (the HCI stands for Human Computer Interaction) does
interdisciplinary work with psychologists.

SUNY Stony Brook: NLP Lab
http://www.cs.sunysb.edu/~nlp/index.html

SUNY Stony Brook: NLP Lab: about
http://www.cs.sunysb.edu/~nlp/about.html

leapinglizard

I´m only rating this as 4 stars because I felt the researcher had more
domain knowledge in the field of discourse analysis to offer than in
the knowledge of psychotherapy, which was unfortunate, but
understandable.  I wish he or she had consulted with a domain expert
in psychotherapy to collaborate to provide a more well rounded answer,
however the information he did provide about linguistics and computer
science was extremely helpful, and am very glad to have it.