Legal Texts Summarization by Exploration of the
Thematic Structures and Argumentative Roles
Atefeh Farzindar and Guy Lapalme
RALI, D´epartement d’Informatique et Recherche Op´erationnelle
Universit´e de Montr´eal, Qu´ebec, Canada, H3C 3J7
ffarzinda,lapalmeg@iro.umontreal.ca
Abstract
In this paper we describe our method for the sum-
marization of legal documents helping a legal ex-
pert determine the key ideas of a judgment. Our
approach is based on the exploration of the docu-
ment’s architecture and its thematic structures in or-
der to build a table style summary for improving co-
herency and readability of the text. We present the
components of a system, called LetSum, built with
this approach, its implementation and some prelim-
inary evaluation results.
1 Introduction
The goal of a summary is to give the reader an accu-
rate and complete idea of the contents of the source
(Mani, 2001). In this research, we focused on a
problem referred to as legal text summarization.
As ever larger amounts of legal documents become
available electronically, interest in automatic sum-
marization has continued to grow in recent years. In
this paper, we present our method for producing a
very short text from a long legal document (a record
of the proceedings of federal courts in Canada) and
present it as a table style summary. The goal of
this project is to develop a system to create a sum-
mary for the needs of lawyers, judges and experts
in the legal domain. Our approach investigates the
extraction of the most important units based on the
identification of thematic structures of the document
and the determination of semantic roles of the tex-
tual units in the judgment (Farzindar, 2004). The
remainder of the paper is organized as follows. Sec-
tion 2 introduces the motivation of the research and
the context of the work. Section 3 reports on the
results of our analysis of a corpus of legal abstracts
written by professional abstractors. Section 4 de-
scribes our method for the exploration of document
architecture and the components of the system that
we have developed to produce a summary. Section 5
presents some related work in this domain. Section
6 concludes the paper and presents some prelimi-
nary evaluation results for the components of our
system.
2 Context of the Work
In Canada, the Canadian Legal Information Insti-
tute project (CANLII) aims at gathering legisla-
tive and judicial texts, as well as legal commen-
taries, from federal, provincial and territorial ju-
risdictions in order to make primary sources of
Canadian law accessible for free on the Internet
(http://www.canlii.org). The large vol-
ume of legal information in electronic form creates
a need for the creation and production of powerful
computational tools in order to extract relevant in-
formation in a condensed form.
But why are we interested in the processing of
previous legal decisions and in their summaries?
First, because a court order generally gives a so-
lution to a legal problem between two or several
parties. The decision also contains the reasons
which justify the solution and constitute a law ju-
risprudence precedent from which it is possible to
extract a legal rule that can be applied to simi-
lar cases. To find a solution to a legal problem
not directly indicated in the law, lawyers look for
precedents of similar cases. For a single query in
a data base of law reports, we often receive hun-
dreds of documents that are very long to study for
which legal experts and law students request sum-
maries. In Quebec REJB (R´epertoire ´electronique
de jurisprudence du Barreau) and SOQUIJ (Soci´et´e
qu´eb´ecoise d’information juridique) are two orga-
nizations which provide manual summaries for le-
gal resources, but the human time and expertise re-
quired makes their services very expensive. For ex-
ample the price of only one summary with its full
text, provided by SOQUIJ is 7.50 $ can. Some legal
information systems have been developed by private
companies like QuickLaw in Canada and WEST-
LAW and LEXIS in the United States, however no
existing system completely satisfies the specific re-
quirements of this field.
One reason for the difficulty of this work is the
complexity of the domain: specific vocabularies of
Between:
JASPER NATIONAL PARK Applicants and THE ATTORNEY GENERAL OF CANADA Respondent,
Docket: T-1557-98
Judgment Professional abstract Role
[1] This application for judicial review arises
out of a decision (the Decision) announced on
or about the 30th of June 1998 by the Minister
of Canadian Heritage (the Minister) to close
the Maligne River (the River) in Jasper Na-
tional Park to all boating activity, beginning
in 1999.
Judicial review of Minister of Canadian
Heritage’s decision to close Maligne River
in Jasper National Park to all boating activ-
ity beginning in 1999 to protect habitat of
harlequin ducks.
INTRO-
DUCTION
[7] The applicants offer commercial rafting
trips to Park visitors in this area each year
from mid-June to sometime in September.
Applicants offer commercial rafting trips on
River.
CONTEXT
[10] Consequently, a further environmental
assessment regarding commercial rafting on
the Maligne River was prepared in 1991. The
assessment indicated that rafting activity had
expanded since 1986, with an adverse impact
on Harlequin ducks along the Maligne River.
1991 environmental assessment indicating
rafting having adverse impact on harlequin
ducks along river.
CONTEXT
Table 1: Alignment of the units of the original judgment with the professional abstract
the legal domain and legal interpretations of expres-
sions produce many ambiguities. For example, the
word sentence can have two very different mean-
ings: one is a sequence of words and the other is a
more particular meaning in law, the decision as to
what punishment is to be imposed. Similarly dis-
position which means nature, effort, mental attitude
or property but in legal terms it means the final part
of a judgement indicating the nature of a decision:
acceptation of a inquiry or dismission.
Most previous systems of automatic summariza-
tion are limited to newspaper articles and scien-
tific articles (Saggion and Lapalme, 2002). There
are important differences between news style and
the legal language: statistics of words, probability
of selection of textual units, position of paragraphs
and sentences, words of title and lexical chains rela-
tions between words of the title and the key ideas of
the text, relations between sentences and paragraphs
and structures of the text.
For judgments, we show that we can identify dis-
cursive structures for the different parts of the deci-
sion and assign some argumentative roles to them.
Newspapers articles often repeat the most important
message but, in law, important information may ap-
pear only once. The processing of a legal document
requires detailed attention and it is not straight for-
ward to adapt the techniques developed for other
types of document to the legal domain.
3 Observations from a Corpus
3.1 Composition
Our corpus contains 3500 judgments of the Federal
Court of Canada, which are available in HTML on
http://www.canlii.org/ca/cas/fct/.
We analyzed manually 50 judgments in English and
15 judgments in French as well as their summaries
written by professional legal abstractors. The
average size of the documents that are input to
our system are judgments between 500 and 4000
words long (2 to 8 pages), which form 80% of all
3500 judgments; 10% of the documents having
less than 500 words (about one page) and so
they do not need a summary. Only 10% of the
decisions have more than 4000 words. Contrary
to some existing systems (Moens et al., 1999) that
focus only on limited types of judgments, such
as criminal cases, our research deals with many
categories of texts such as: Access to information,
Administrative law, Air law, Broadcasting, Com-
petition, Constitutional law, Copyright, Customs
and Excise - Customs Act, Environment, Evidence,
Human rights, Maritime law, Official languages,
Penitentiaries, Unemployment insurance and etc.
3.2 Structure of Legal Judgments
During our corpus analysis, we compared model
summaries written by humans with the texts of
the original judgments. We have identified the
organisational architecture of a typical judgment.
Thematic structures Content Judgment Summary
DECISION DATA Name of the jurisdiction,
place of the hearing,
date of the decision,
identity of the author,
names of parties,
title of proceeding and
Authority and doctrine
INTRODUCTION Who? did what? to whom? 5 % 12 %
CONTEXT Facts in chronological order or by description 24 % 20 %
JURIDICAL ANALYSIS Comments by the judge, finding of facts and
application of the law
67 % 60 %
CONCLUSION Final decision of the court 4 % 8 %
Table 2: Table of summary shows the thematic structures in a jugement and percentage of the contribution
of each thematic structure in source judgment and its human made summary
!"#$%&’(
)#*$#+&%&’,+
)#-#(&’,+
./,01(&’,+
20#+&’3(%&’,+4,5
610*$#+&
,/*%+’7%&’,+4
8’+04&"#
/#-#9%+&41+’&744
:;&/%(&4&"#
"’*"#7&47(,/#0
1+’&744
817’,+4&"#
#;&/%(&#04<%/&7
:-’$’+%&’,+4,5
1+’$<,/&%+&
#-#$#+&7
!%=-#47&>-#
71$$%/>
8’-&#/’+*
?,’7#
@#01(&’,+4
A#*%-
B,(1$#+&
:-’$’+%&’,+4,5
C1,&%&’,+74
B#&#/$’+%&’,+4,5
)#$%+&’(4@,-#7
Figure 1: The procedural steps for generating of table style summary
The paragraphs that address the same subject are
grouped as members of a block. We annotated the
blocks with a label describing their semantic roles.
We also manually annotated citations which are tex-
tual units (sentence or paragraph) quoted by the
judge as reference, for example an article of law
or other jurisprudence. The citations account for a
large part of the text of the judgment, but they are
not considered relevant for the summary, therefore
these segments will be eliminated during the infor-
mation filtering stage.
The textual units considered as important by the
professional abstractors were aligned manually with
one or more elements of the source text. Table 1
shows an example of an alignment between a human
summary and the original judgment. We look for a
match between the information considered impor-
tant in the professional abstract and the information
in the source documents. Our observation shows
that, for producing a summary, a professional ab-
stractor mainly relies on the manual extraction of
important units while conforming to general guide-
lines. The collection of these selected units forms a
summary.
During this analysis, we observed that texts
of jurisprudence are organized according to a
macrostructure and contain various levels of infor-
mation, independently of the category of judgment.
Proposed guidelines by Judge Mailhot of the Court
of Appeal of Quebec (Mailhot, 1998) and (Branting
et al., 1997) on legal judgments support this idea
that it is possible to define organisational structures
for decisions. Jurisprudence is organized by the dis-
course itself, which makes it possible to segment the
texts thematically.
Textual units dealing with the same subject form
a thematic segment set. In this context, we distin-
guish the layered thematic segments, which divide
the legal decisions into different discursive struc-
tures. The identification of these structures sepa-
rates the key ideas from the details of a judgment
and improves readability and coherency in the sum-
mary. We will present the argumentative roles of
each level of discourse, and their importance in
the judgment from the point of view of the key
and principal ideas. Table 2 shows the structure
of a jurisprudence and its different discourse lev-
els. Therefore, in the presentation of a final sum-
mary, we propose to preserve this organization of
the structures of the text in order to build a table
style summary with five themes:
DECISION DATA contains the name of the jurisdic-
tion, the place of the hearing, the date of the de-
cision, the identity of the author, names of par-
ties, title of proceeding, authority and doctrine.
It groups all the basic preliminary information
which is needed for planning the decision.
INTRODUCTION describes the situation before the
court and answers these questions: who are the
parties? what did they do to whom?
CONTEXT explains the facts in chronological or-
der, or by description. It recomposes the story
from the facts and events between the par-
ties and findings of credibility on the disputed
facts.
JURIDICAL ANALYSIS describes the comments
of the judge and finding of facts, and the ap-
plication of the law to the facts as found. For
the legal expert this section of judgment is the
most important part because it gives a solution
to the problem of the parties and leads the judg-
ment to a conclusion.
CONCLUSION expresses the disposition which is
the final part of a decision containing the infor-
mation about what is decided by the court. For
example, it specifies if the person is discharged
or not or the cost for a party.
During our corpus analysis, we computed the
distribution of the information (number of words
shown in Table 2) in each level of thematic structure
of the judgment. The average length of a judgment
is 3500 words and 350 words for its summary i.e. a
compression rate of about 10%.
4 Method for Producing Table Style
Summary
Our approach for producing the summary first iden-
tifies thematic structures and argumentative roles in
the document. We extract the relevant sentences and
present them as a table style summary. Showing the
information considered important which could help
the user read and navigate easily between the sum-
mary and the source judgment. For each sentence
of the summary, the user can determine the theme
by looking at its rhetorical role. If a sentence seems
more important for a user and more information is
needed about this topic, the complete thematic seg-
ment containing the selected sentence could be pre-
sented. The summary is built in four phases (Fig-
ure 1): thematic segmentation, filtering of less im-
portant units such as citations of law articles, selec-
tion of relevant textual units and production of the
summary within the size limit of the abstract.
The implementation of our approach is a system
called LetSum (Legal text Summarizer), which has
been developed in Java and Perl. Input to the sys-
tem is a legal judgment in English. To determine the
Part-of-Speech tags, the tagger described by (Hep-
ple, 2000) is used. The semantic grammars and
rules are developed in JAPE language (Java Anno-
tations Pattern Engine) and executed by a GATE
transducer (Cunningham et al., 2002).
4.1 Components of LetSum
Thematic segmentation for which we performed
some experiments with two statistic segmenters:
one described by Hearst for the TexTiling system
(Hearst, 1994) and the C99 segmenter described by
Choi (Choi, 2000), both of which apply a clustering
function on a document to find classes divided by
theme. But because the results of these numerical
segmenters were not satisfactory enough to find the
thematic structures of the legal judgments, we de-
cided to develop a segmentation process based on
the specific knowledge of the legal field.
Category of section title Linguistic markers Examples of section title
Begin of the judgment decision, judgment, reason,
order
Reasons for order, Reasons for
judgment and order
INTRODUCTION introduction, summary Introduction, Summary
CONTEXT facts, background The factual background, Agreed
statement of facts
JURIDICAL ANALYSIS analysis, decision, discussion Analysis and Decision of the court
CONCLUSION conclusion, disposiotion, cost Conclusion and Costs
Table 3: The linguistic markers in section titles
Each thematical segment can be associated with
an argumentative role in the judgment based on the
following information: the presence of significant
section titles (Table 3 shows categories and features
of the section titles), the absolute and relative posi-
tions of a segment, the identification of direct or nar-
rative style (as the border of CONTEXT and JURIDI-
CAL ANALYSIS segments), certain linguistic mark-
ers.
The linguistic markers used for each thematic
segment are organized as follows:
CONTEXT introduces the parties with the verb to
be (eg. the application is company X), describes the
application request like: advise, indicate, request
and explains the situation in the past tense and nar-
ration form.
In JURIDICAL ANALYSIS, the judge gives his ex-
planation on the subject thus the style of expression
is direct such as: I, we, this court, the cue phrases
(Paice, 1981) like: In reviewing the sections No. of
the Act, Pursuant to section No., As I have stated, In
the present case, The case at bar is.
In CONCLUSION the classes of verbs are: note,
accept, summarise, scrutinize, think, say, satisfy,
discuss, conclude, find, believe, reach, persuade,
agree, indicate, review, the concepts such as: opin-
ion, conclusion, summary, because, cost, action, the
cue phrases: in the case at bar, for all the above
reasons, in my view, my review of, in view of the evi-
dence, finally, thus, consequently, in the result. This
segment contains the final result of court decision
using phrases such as: The motion is dismissed, the
application must be granted. The important verbs
are: allow, deny, dismiss, grant, refuse.
Filtering identifies parts of the text which can be
eliminated, without losing relevant information for
the summary. In a judgment, the citation units (sen-
tence or paragraph) occupy a large volume in the
text, up to 30%, of the judgment, whereas their con-
tents are less important for the summary. This is
why we remove citations inside blocks of thematic
segments. We thus filter two categories of segments:
submissions and arguments that report the points of
view of the parties in the litigation and citations re-
lated for previous issues or references to applicable
legislation. In the case of eliminating a citation of
a legislation (eg. law’s article), we save the refer-
ence of the citation in DECISION DATA in the field
of authority and doctrine.
The identification of citations is based on two
types of markers: direct and indirect. A direct
marker is one of the linguistic indicators that we
classified into three classes: verbs, concepts (noun,
adverb, adjective) and complementary indications.
Examples of verbs of citation are: conclude, define,
indicate, provide, read, reference, refer, say, state,
summarize. Examples of the concepts are: follow-
ing, section, subsection, page, paragraph, pursuant.
Complementary indications include numbers, cer-
tain preposition, relative clauses and typographic
marks (colon, quotation marks).
The indirect citations are the neighboring units of
a quoted phrase. For example, in Table 4 a citation
is shown. For detecting CITATION segment units
such as paragraph 78(1), which reads as follows:
are identified using direct markers (shown here in
bold) but surrounding textual units with numbers
are also quotations. We thus developed a linear inte-
gration identification mechanism for sentences fol-
lowing a quoted sentence for determining a group
of citations.
Selection builds a list of the best candidate units
for each structural level of the summary. LetSum
computes a score for each sentence in the judgment
based on heuristic functions related to the following
information: position of the paragraphs in the doc-
ument, position of the paragraphs in the thematic
segment, position of the sentences in the paragraph,
distribution of the words in document and corpus
(tf  idf ). Depending on the given information in
each layered segment, we have identified some cue
words and linguistic markers. The thematic segment
can change the value of linguistic indicators. For ex-
ample, the phrase application is dismissed that can
be considered as a important feature in the CON-
CLUSION might not have the same value in CON-
TEXT segment. At the end of this stage, the pas-
sages with the highest resulting scores are sorted to
determine the most relevant ones.
Production of the final summary in which the
selected sentences are normalized and displayed in
tabular format. The final summary is about 10%
of source document. The elimination of the unim-
portant sentences takes into account length statistics
presented in Table 2. In the INTRODUCTION seg-
ment, units with the highest score are kept within
10% of the size of summary. In the CONTEXT seg-
ment, the selected units occupy 24% of the sum-
mary length. The contribution of the JURIDICAL
ANALYSIS segment is 60% and the units with the
role CONCLUSION occupy 6% of the summary.
4.2 Current State of LetSum
Table 4 shows an example of the output after the ex-
ecution of the Selection module of LetSum (mod-
ules of Figure 1 up to the horizontal line) applied
on a judgment of Federal Court of Canada (2468
words). Thematic segmentation module has di-
vided the text into structural blocks according to
the rhetorical roles (given to the left of braces in
Table 4). The Filtering module removes citation
blocks and its enumerated quoted paragraphs (e.g.
paragraph (15) in tablet). Selection module chooses
total relevant textual units (shown in bold in Table 4)
in each thematic segment. The units are selected
according to their argumentative role in the judge-
ment. Here the length of all extracted units is 313
words.
Preliminary evaluations of components of Let-
Sum are very promising; we obtained 0.90 F-
measure for thematic segmentation and 0.97 F-
measure for filtering stage (detection of 57 quoted
segment correctly on 60).
From this information, the Production module
(currently being implemented) could concatenate
textual units with some grammatical modification to
produce a short summary.
5 Related Research
LetSum is the one of the few systems developed
specifically for the summarization of legal docu-
ments. All of these approaches attest the impor-
tance of the exploration of thematic structures in le-
gal documents.
The FLEXICON project (Smith and Deedman,
1987) generates a summary of legal cases by us-
ing information retrieval based on location heuris-
tics, occurrence frequency of index terms and the
use of indicator phrases. A term extraction module
that recognizes concepts, case citations, statute ci-
tations and fact phrases leads to a document profile.
This project was developed for the decision reports
of Canadian courts, which are similar to our corpus.
SALOMON (Moens et al., 1999) automatically
extracts informative paragraphs of text from Belgian
legal cases. In this project a double methodology
was used. First, the case category, the case struc-
ture and irrelevant text units are identified based
on a knowledge base represented as a text gram-
mar. Consequently, general data and legal foun-
dations concerning the essence of the case are ex-
tracted. Secondly, the system extracts informative
text units of the alleged offences and of the opinion
of the court based on the selection of representative
objects.
More recently, SUM (Grover et al., 2003) exam-
ined the use of rhetorical and discourse structure in
level of the sentence of legal cases for finding the
main verbes. The methodology is based on (Teufel
and Moens, 2002) where sentences are classified ac-
cording to their argumentative role.
These studies have shown the interest of summa-
rization in a specialized domain such as legal texts
but none of these systems was implemented in an
environment such as CANLII which has to deal with
thousands of texts and produce summaries for each.
6 Conclusion
In this paper, we have presented our approach for
dealing with automatic summarization techniques.
This work refers to the problem of processing of a
huge volume of electronic documents in the legal
field which becomes more and more difficult to ac-
cess. Our method is based on the extraction of rele-
vant units in the source judgment by identifying the
discourse structures and determining the semantic
roles of thematic segments in the document. The
presentation of the summary is in a tabular form di-
vided by the following thematic structures: DECI-
SION DATA, INTRODUCTION, CONTEXT, JURIDI-
CAL ANALYSIS and CONCLUSION. The generation
of summary is done in four steps: thematic segmen-
tation to detect the document structures, filtering to
eliminate unimportant quotations and noises, selec-
tion of the candidate units and production of table
style summary. The system is currently being fi-
nalized and preliminary evaluation results are very
promising.
7 Acknowledgements
We would like to thanks LexUM group of le-
gal information-processing laboratory of the Public
DECISION DATA
8
>>>
><
>>>
>:
Name of the jurisdiction: Federal Court of Canada, Place of the hearing: Ottawa
Date of the decision: 31/12/97, Identity of the author: J.E. Dub´e
Names of parties: Commissioner of official languages of canad, Applicant
- and - Air Canada, Respondent
Title of proceeding: Official languages, Docket number: T-1989-96
Authority and doctrine : Official Languages Act, R.S.C., 1985 (4th Supp.), c. 31
INTRODUCTION
8
>><
>>:
(1) An order was made by this Court on February 4, 1997 authorizing the respondent (Air Canada) to
raise preliminary objections to the notice of an originating motion filed by the applicant (the Commis-
sioner). As a result, this motion filed by Air Canada on March 18, 1997 raises six alternative prelim-
inary objections asking the Court to strike out in part the motion made by the Commissioner on
September 6, 1996 under section 78 of the Official Languages Act.
CONTEXT
8
>>>
>>>
>>>
<
>>>
>>>
>>>
>:
1. Facts
(2) The Commissioner’s originating motion, which was filed with the consent of the com-
plainant Paul Comeau, concerns Air Canada’s failure to provide ground services in the
French language at the Halifax airport. The Commissioner asks this Court to declare that
there is a significant demand for services in French in Air Canada”s office at the Halifax airport
and that Air Canada is failing to discharge its duties under Part IV of the Act. Part IV estab-
lishes language-related duties for communications with and services to the public, including the
travelling public, where there is significant demand.
(3) The Commissioner’s motion is filed by the complainant Paul Comeau.
...
CITATION
8
>>>
>>>
<
>>>
>>>
:
(15) The point of departure is paragraph 78(1), which reads as follows:
78. (1) The Commissioner may
(a) within the time limits prescribed by paragraph 77(2)( a) or ( b), apply to the Court for
a remedy under this Part in relation to a complaint investigated by the Commissioner if the
Commissioner has the consent of the complainant.
(b) appear before the Court on behalf of any person who has applied under section 77 for a
remedy under this Part; or
(c) with leave of the Court, appear as a party to any proceedings under this Part.
ANALYSIS
8>
>><
>>>
:
(16) Air Canada’s position is therefore that the Commissioner may only apply for a rem-
edy limited to facts relating to a specific complaint, the investigation of that complaint
and the resulting reports and recommendations. In my view, this interpretation is too
narrow and is inconsistent with the general objectives of the Act and its remedial and
quasi-constitutional nature.
...
CONCLUSION
8
>>>
><
>>>
>:
7. Conclusion
(29) Thus, to ensure that the judge presiding at the hearing on the merits can correctly assess the
situation in light of all the material evidence, no reference or evidence filed by the Commissioner
in the three affidavits mentioned above should be struck out.
(30) This motion to strike by Air Canada with respect to the preliminary objections must accord-
ingly be dismissed.
Table 4: Output produced by the LetSum’s modules: Thematic segmentation, Filtering and Selection.
Source judgment is divided into thematic blocks associated with rhetorical roles, citation block will be
removed in the filtering phase and textual units (shown in bold) have been selected as relevant.
Law Research Center at the University of Montreal
for their valuable suggestions. This project sup-
ported by Public Law Research Center and Natu-
ral Sciences and Engineering Research Council of
Canada.

References
L. Karl Branting, Charles B. Callaway, Bradford W.
Mott, and James C. Lester. 1997. A frame-
work for self-explaining legal documents. In
Proceedings of the Sixth International Confer-
ence on Artificial Intelligence and Law (ICAIL-
97), pages 72–81, University of Melbourne, Mel-
bourne, Australia, June 30-July 3.
Freddy Choi. 2000. Advances in domain indepen-
dent linear text segmentation. In Proceding of
the 1 st North American Chapter of the Associa-
tion for Computational Linguistics, pages 26–33,
Seattle, Washington.
H. Cunningham, D. Maynard, K. Bontcheva, and
V. Tablan. 2002. Gate: A framework and graphi-
cal development environment for robust nlp tools
and applications. In Proceedings of the 40th An-
niversary Meeting of the Association for Compu-
tational Linguistics (ACL’02), Philadelphia, July.
Atefeh Farzindar. 2004. D´eveloppement d’un
syst`eme de r´esum´e automatique de textes ju-
ridiques. In TALN-RECITAL’2004, pages 39–44,
F`es, Maroc, 19-22 April.
Claire Grover, Ben Hachey, and Chris Korycinski.
2003. Summarising legal texts: Sentential tense
and argumentative roles. In Dragomir Radev and
Simone Teufel, editors, HLT-NAACL 2003 Work-
shop: Text Summarization (DUC03), pages 33–
40, Edmonton, Alberta, Canada, May 31 - June
1.
Marti A. Hearst. 1994. Multi-paragraph segmenta-
tion of expository text. In the 32nd Meeting of the
Association for Computational Linguistics, Los
Cruces, NM, June.
Mark Hepple. 2000. Independence and commit-
ment: Assumptions for rapid training and execu-
tion of rule-based part-of-speech taggers. In the
38th Annual Meeting of the Association for Com-
putational Linguistics (ACL-2000), pages 278–
285, October.
Louise Mailhot. 1998. Decisions, Decisions: a
handbook for judicial writing. Editions Yvon
Blais, Qu´ebec, Canada.
Inderjeet Mani. 2001. Automatic Text Summariza-
tion. John Benjamins Publishing Company.
Marie-Francine Moens, C. Uyttendaele, and J. Du-
mortier. 1999. Abstracting of legal cases: the
potential of clustering based on the selection of
representative objects. Journal of the American
Society for Information Science, 50(2):151–161.
Chris D. Paice. 1981. The automatic generation of
literary abstracts: An approach based on identi-
fication of self-indicating phrases. In O. R. Nor-
man, S. E. Robertson, C. J. van Rijsbergen, and
P. W. Williams, editors, Information Retrieval
Research, London: Butterworth.
Horacio Saggion and Guy Lapalme. 2002. Gener-
ating indicative-informative summaries with su-
mum. Computational Linguistics, 28(4).
J. C. Smith and Cal Deedman. 1987. The applica-
tion of expert systems technology to case-based
law. ICAIL, pages 84–93.
Simone Teufel and Marc Moens. 2002. Summaris-
ing scientific articles - experiments with rele-
vance and rhetorical status. Computational Lin-
guistics, 28(4):409–445.
