Lexically-Based Terminology Structuring: Some Inherent Limits
Natalia Grabar and Pierre Zweigenbaum
STIM/DSI, Assistance Publique – Hôpitaux de Paris
& Département de Biomathématiques, Université Paris 6
{ngr,pz}@biomath.jussieu.fr
http://www.biomath.jussieu.fr/˜{ngr,pz}
Abstract
Terminology structuring has been the subject of
much work in the context of terms extracted from
corpora: given a set of terms, obtained from an ex-
isting resource or extracted from a corpus, identi-
fying hierarchical (or other types of) relations be-
tween these terms. The present paper focusses on
terminology structuring by lexical methods, which
match terms on the basis on their content words,
taking morphological variants into account. Exper-
iments are done on a ‘flat’ list of terms obtained
from an originally hierarchically-structured termi-
nology: the French version of the US National
Library of Medicine MeSH thesaurus. We com-
pare the lexically-induced relations with the original
MeSH relations: after a quantitative evaluation of
their congruence through recall and precision met-
rics, we perform a qualitative, human analysis of the
‘new’ relations not present in the MeSH. This anal-
ysis shows, on the one hand, the limits of the lex-
ical structuring method. On the other hand, it also
reveals some specific structuring choices and nam-
ing conventions made by the MeSH designers, and
emphasizes ontological commitments that cannot be
left to automatic structuring.
1 Background
Terminology structuring, i.e., organizing a set of
terms through semantic relations, is one of the dif-
ficult issues that have to be addressed when build-
ing terminological resources. These relations in-
clude subsumption or hyperonymy (the is-a re-
lation), meronymy (part-of and its variants), as
well as other, diverse relations, sometimes called
‘transversal’ (e.g., cause, or the general see also).
Various methods have been proposed
to discover relations between terms (see
Jacquemin and Bourigault (2002) for a review).
We divide them into internal and external meth-
ods, in the same way as McDonald (1993)
for proper names. Internal methods look
at the constituency of terms, and compare
terms based on the words they contain. Term
matching can rely directly on raw word forms
(Bodenreider et al., 2001), on morphological
variants (Jacquemin and Tzoukermann, 1999),
on syntactic structure (Bourigault, 1994;
Jacquemin and Tzoukermann, 1999) or on se-
mantic variants (synonyms, hyperonyms, etc.)
(Hamon et al., 1998). External methods take
advantage of the context in which terms occur:
they examine the behavior of terms in corpora.
Distributional methods group terms that occur
in similar contexts (Grefenstette, 1994). The
detection of appropriate syntactic patterns of
cooccurrence is another method to uncover re-
lations between terms in corpora (Hearst, 1992;
Séguéla and Aussenac, 1999).
In previous work we applied lexical methods to
identify relations between terms on the basis on
their content words, taking morphological variants
into account. Our goal was then to assess the feasi-
bility of such structuring by studying it on an exist-
ing, hierarchically structured terminology. Ignoring
this existing structure and starting from the set of its
terms, we attempt to discover hierarchical term-to-
term links and compare them with the preexisting
relations.
Our goal in the present paper is to analyze ‘new’
relations. ‘New’ means that these induced relations
are not present in the original hierarchical structure
of the MeSH thesaurus; they might nevertheless re-
flect useful links. Performing this analysis allows us
to propose a more precise evaluation of the methods
and their results and to point out some inherent lim-
its.
After the exposition of the data we used in our
experiments (section 2), we present methods (sec-
tion 3) for generating hierarchical links between
terms through the study of lexical inclusion and for
evaluating their quality with appropriate recall and
precision metrics. We then present the analysis of
some ‘new’ induced relations and attempt to pro-
pose a typology of term dependency in these rela-
tions (section 4). We finally discuss the limits of
lexical methods for the structuring task (section 5).
2 The MeSH biomedical thesaurus, and
associated morphological knowledge
We first present the existing hierarchically struc-
tured thesaurus, a ‘stop word’ list and morpholog-
ical knowledge involved in the present work.
2.1 The MeSH biomedical thesaurus
The Medical Subject Headings (MeSH,
NLM (2001a)) is one of the main international
medical terminologies (see, e.g., Cimino (1996)
for a presentation of medical terminologies). It is
a thesaurus specifically designed for information
retrieval in the biomedical domain. The MeSH is
used to index the international biomedical literature
in the Medline bibliographic database. The French
version of the MeSH (INSERM, 2000) contains
a translation of these terms (19,638 terms) plus
synonyms. It happens to be written in unaccented,
uppercase letters. Both the American and French
MeSH can be found in the UMLS Metathesaurus
(NLM, 2001b), which can be obtained through a
convention with the National Library of Medicine.
The concept names (main headings) which the
MeSH contains have been designed to reflect their
broad meanings and to facilitate their use by hu-
man indexers and librarians. In that, they follow a
tradition in information sciences, and are not nec-
essarily the expressions used in naturally occurring
biomedical documents. The MeSH can be consid-
ered as a fine-grained thesaurus: concepts are cho-
sen to insure a good coverage of the biomedical do-
main (Zweigenbaum, 1999).
As many other medical terminologies, the MeSH
has a hierarchical structure: ‘narrower’ concepts
(children) are related to ‘broader’ concepts (par-
ents). This both covers the usual is-a relation and
partitive relations (part-of, conceptual-part-of and
process-of ). The MeSH also includes see-also re-
lations, which we do not take into account in the
present experiments. This structure has also been
designed in the aim to be intellectually accessi-
ble to users: an indexer must be able to assign a
given concept to an article and a clinician must be
able to find a given concept in the tree hierarchy
(Nelson et al., 2001). To conclude, the MeSH team
aims to organize it in a clear and intuitive manner,
both for concept naming and concept placement.
The version of the French MeSH we used in these
experiments contains 19,638 terms, 26,094 direct
child-to-parent links and (under transitive closure)
95,815 direct or indirect child-to-ancestor links.
2.2 Stop word list
The aim of using a ‘stop word’ list is to remove from
term comparison very frequent words which are
considered not to be content-bearing, hence ‘non-
significant’ for terminology structuring. We used
in this experiment a short stop word list (15 word
forms). It contains the few frequent grammatical
words, such as articles and prepositions, that occur
in MeSH terms.
2.3 Morphological knowledge
The morphological knowledge involved consists of
lemma/derived-word or lemma/inflected form pairs
where the first is the ‘normalized’ form and the sec-
ond a ‘variant’ form.
Inflection produces the various forms of a given
word such as plural, feminine or the multiple forms
of a verb according to person, tense, etc.: inter-
vention – interventions, acid – acids. We per-
form the reverse process (lemmatization), reducing
an inflected form to its lemma (canonical form).
We worked with two alternate lexicons. The first
one is based on a general French lexicon (ABU,
abu.cnam.fr/DICO) which we have augmented with
pairs obtained from medical corpora processed
through a tagger/lemmatizer (in cardiology, hema-
tology, intensive care, and drug monographs): it to-
tals 219,759 pairs (where the inflected form is dif-
ferent from the lemma). The second lexicon, more
specialized and tuned to the vocabulary in medi-
cal terminologies, is the result of applying rules ac-
quired in previous work from two other medical ter-
minologies (ICD-10 and SNOMED) to the vocab-
ulary in the MeSH, ICD-10 and SNOMED (total:
2,889 pairs).
Derivation produces, e.g., the adjectival form of
a noun (noun aorta a0 adjective aortic), the nom-
inal form of a verb (verb intervene a0 noun inter-
vention), or the adverbial form of an adjective (ad-
jective human a0 adverb humanely). We perform
linguistically-motivated stemming to reduce a de-
rived word to its base word. For derivation, we also
used resources acquired in previous work which,
once combined with inflection pairs, results in 4,517
pairs.
Compounding, which combines several radicals,
often of Greek or Latin origin, to obtain complex
words (e.g., aorta + coronary yields aortocoro-
nary), has not been used because we do not have
a reliable procedure to segment a compound into its
component morphemes.
3 Acquiring links through lexical
inclusion of terms
The present work induces hierarchical relations be-
tween terms when the constituent words of one term
lexically include those of the second term (sec-
tion 3.1). When comparing these relations with
those that preexist in the MeSH, precision can reach
29.3% and recall 13.7% (section 3.2). We focus
here on the analysis of the relations that are not
found in the MeSH (section 3.3), which we develop
in the next section (section 4).
3.1 Lexical inclusion
The method we use here for inducing hierarchical
relations between terms is basically a test of lexical
inclusion: we check whether a term a0 (parent) is
‘included’ in another term a1 (child), i.e., whether
all words in a0 occur in a1 . We assume that this type
of inclusion is a clue of a hierarchical relation be-
tween terms, as in acides gras / acides gras indis-
pensables (fatty acids / fatty acids, essential).
To detect this type of relation, we test whether
all the content words of a0 occur in a1 . We do this
on segmented terms with a gradually increasing nor-
malization on word forms. Basic normalizations are
performed first: conversion to lower case, removal
of punctuation, of numbers and of ‘stop words’.
Subsequent normalizations rely on morphological
ressources: lemmatization (with the two alternate
inflectional lexicons) and stemming with a deriva-
tional lexicon. Terms are indexed by their words to
speed up the computation of term inclusion over all
term pairs of the whole MeSH thesaurus.
3.2 Application to MeSH and quantification
This structuring method has been applied to the flat
list of 19,638 terms of the MeSH thesaurus. As ex-
pected, the number of links induced between terms
increases when applying inflectional normalization
and again with derivational normalization.
We evaluated the quality of the links obtained
with this approach by comparing them automati-
cally with the original structure of the MeSH and
computing recall and precision metrics. We sum-
marize here the main results; a detailed evaluation
can be found in (Grabar and Zweigenbaum, 2002).
Depending on the normalization, up to 29.3% of
the links found are correct (precision), and up to
13.7% of the direct MeSH links are found by lex-
ical inclusion (recall). We also examined whether
each term was correctly placed under one of its an-
cestors: this was true for up to 26% of the terms
(recall); and the placement advices were correct in
up to 58% of the cases (precision). The recall of
links increases when applying more complete mor-
phological knowledge (inflection then derivation).
The evolution of precision is opposite: injection of
more extensive morphological knowledge (deriva-
tion vs inflection) leads to taking more ‘chances’ for
generating links between terms: the precision with
no normalization (raw results) is 29.3% vs 22.5%
when using all normalizations (lem-stem-med). De-
pending on the type of normalization, the best pre-
cision obtained for links is 43%.
3.3 Human analysis of ‘new’ relations
The evaluations presented in the previous section
quantify the match between the induced relations
and existing MeSH relations. However, they give
no explanation for the fact that 70% of the induced
relations are not considered relevant by the MeSH.
This is what we study in the remainder of this paper:
why these terms are not hierarchically related in the
MeSH, and what kinds of relations exist between
them.
According to the position of the words of the
‘parent’ term in the ‘child’ term, we divide the
extra-MeSH relations into three sets: a2a4a3a6a5 the par-
ent concept is at the head position in the child con-
cept: absorption/absorption intestinale; a2a8a7a9a5 the par-
ent concept is at the tail (expansion) position in the
child concept: abdomen/tumeur abdomen; a2a8a10a9a5 other
types of positions. Each set of relations is sam-
pled by randomly selecting a 20% subset, both with-
out normalization (raw) and with inflectional and
derivational normalizations (med-lem-stem). Ta-
ble 1 presents the number of analyzed relations (to-
tal = 194).
Normalizations Head Expan. Other
raw 22 31 14
lem-stem-med 37 57 33
Table 1: Relations to analyze: sample sizes.
4 An analysis of new, lexically-induced
relations
We first examine the issues encountered when try-
ing to identify the head of each term (section 4.1),
then review in turn each analyzed subset: head (sec-
tion 4.2), expansion (section 4.3) and other relations
(section 4.4).
4.1 Finding the head
In French, the semantic head of a noun phrase is
usually located at the beginning of this phrase (this
contrasts with English, where the semantic head is
generally at the end of NPs). Moreover, as is often
the case with terms, MeSH terms do not include de-
terminers, so that the semantic head is usually the
first word here. We therefore rely on a heuristic
for determining ‘head’ and ‘expansion’ subsets: the
head is the first word of the term, and the expansion
is the last word. This is correct most of the time, but
in some cases, the semantic head is positioned at the
end of the term, generally separated with a comma,
a tradition sometimes followed in thesauri:
filoviridae/filoviridae, infections,
leishmania/leishmania tropica, infection,
quinones/quinone reductases,
neurone/neurone moteur, maladie,
syndrome/bouche main pied, syndrome.
These cases must be hand-corrected and distributed
into the following classes.
We also encountered another kind of error, due to
overzealous derivational knowledge:
contracture/contraction musculaire,
biologie/testament biologique,
where contracture (a muscle disease) and con-
traction (normal muscle function) have both been
stemmed to the same base word; the expansion ad-
jective biologique is derived from the noun biologie,
but its sense is generally more specific than biolo-
gie.
4.2 ‘Head’ subset
Let us first discard a case where it seems that we
encountered a translation error. An examination of
the structure of the English MeSH and a search on
Web pages show that in the French MeSH, acide
linoleique alpha should read acide linolenique al-
pha, which is a kind of acide linolenique (and not a
kind of acide linoleique). The induced relation:
acide linoleique/acide linoleique alpha
is therefore incorrect; with the correct spelling, the
lexical inclusion:
acide linolenique/acide linolenique alpha
would reveal a correct hierarchical relation.
4.2.1 The head is not the ‘genus’ of the term
We encountered cases where the whole term did not
have an is-a relation with the head as defined above.
This happens in two types of situations.
The first situation is due to syntactic reasons. In
the following induced relation,
acides amines / acides amines, peptides et pro-
teines,
the larger term is an enumeration, with the sense
of a logical OR. It is therefore the genus term, of
which each of its components (e.g., acides amines)
is a sub-type.
The second situation is due to semantic reasons.
Lexical induction of hierarchical relations assumes
inheritance of the defining features of the genus
term (e.g., a fatty acid, essential is a kind of fatty
acid). However, it is well known that this is not al-
ways true: a plaster cat is not a cat (i.e., a mammal,
etc.). This is sometimes modeled as a type coercion
phenomenon. We found quite a few ‘plaster cats’ in
our terms:
personnalite/personnalite compulsive,
voix/voix oesophagienne.
For instance, personnalite here describes ‘behavior-
response patterns that characterize the individual’,
whereas personnalite compulsive (compulsive per-
sonality disorder) describes a mental disorder. Dis-
orders (or diseases) are different objects than behav-
iors in the MeSH.
4.2.2 The head is ambiguous
This depends on the choice of term names in the ter-
minology (here, the MeSH). Terms like absorption,
investissement, etc., have specific senses that make
them polysemous. To determine a precise sense,
these terms have to be specialized by their contexts:
investissement/investissement (psychanalyse),
absorption/absorption cutanee,
goitre/goitre ovarien
Here, investissement alone (investment) has the fi-
nancial sense, whereas in investissement (psych-
analyse), it has its more generic sense. In a simi-
lar way, absorption has a specific meaning in chem-
istry, and goitre alone is a disorder of the thyroid
gland. These cases are often non-ambiguous in the
original English version of the same terms: for in-
stance, investissement (psychanalyse) (fr) is a trans-
lation of cathexis (en).
A related case occurs when the name of a parent
term is underspecified:
acides/acides pentanoiques,
acne/acne rosacee.
In these examples, acides means inorganic acids1
and acne means acne vulgaris, but the convention
adopted is to use these single words to name the cor-
responding concepts.
4.2.3 Ontological commitment
Finally, some induced links, although absent from
the MeSH, are potentially correct is-a links, but the
designers of the MeSH have made a different mod-
eling choice:
amyotrophies/amyotrophies spinales enfance,
hyperplasie/hyperplasie epitheliale focale,
centre public sante/centre public sante men-
tale,
rectocolite/rectocolite hemorragique,
penicillines/penicilline g.
A general representational choice in the MeSH,
as in some other medical terminologies (e.g.,
SNOMED), is to differentiate on the one hand signs
or symptoms and on the other hand diseases (a
more fully characterized pathological state). This
is the case for amyotrophies and hyperplasie (signs
or symptoms) vs amyotrophies spinales enfance and
hyperplasie epitheliale focale (disease of the ner-
vous system, of the mouth).
For some reason, a centre public sante mentale
(public mental health center) is considered not to
share all the attributes of a general centre public
sante (public health center), which prevents them
from being in a parent-child relationship: they are
only siblings in the MeSH thesaurus.
Penicillines, in the MeSH, have been chosen to
refer to a therapeutic class of drugs (under antibi-
otics, under chemical actions), whereas penicilline
g is considered as a chemical substance.
The structuring involved in these instances re-
flects the ontological commitments of the terminol-
1Note, though, that if inorganic acids was named this way,
it would be impossible to link it by lexical induction to other,
more specific types of inorganic acids.
ogy designers, and cannot be recovered by lexical
inclusion.2
4.3 ‘Expansion’ subset
When a ‘parent’ term is in ‘expansion’ position (end
position) in a ‘child’ term, we assume that the se-
mantic head of the child term is modified; the in-
duced relation is indeed expected not to be is-a.
Some of the main cases found are close to those for
the ‘head’ subset. Among others, we find again enu-
merations (see subsection 4.2.1):
immunodepresseurs / antineoplasiques et im-
munodepresseurs
and syntactic ambiguity (subsection 4.2.2):
oncogene/antigene viral oncogene,
where the word oncogene is a noun in the first term
and an adjective in the second one.
Many of the relations found in the ‘expansion’
subset are partitive:
abdomen/muscle droit abdomen,
amerique centrale/indien amerique centrale,
argent/nitrate argent.
(human body parts, a continent and its peoples, and
chemical substances).
In some instances, a general type of link between
terms can be detected:
caused-by: myxome/virus myxome,
but in most other cases, we have what looks like a
specific thematic relation between a predicate and
its argument:
comportement alimentaire/troubles comporte-
ment alimentaire,
bovin/pneumonie interstitielle atypique bovin,
hopital/capacite lits hopital,
services sante/fermeture service sante,
macrophage/activation macrophage.
Note that some of these expansion relations involve
adjectival derivations of nouns:
cubitus/nerf cubital,
genes/epreuve complementation genetique.
2They might be amenable to distributional methods if their
contexts of occurrence are different enough.
4.4 ‘Other’ subset
In this last subset, the ‘parent’ term can be at any
position in the ‘child’ term other than head or ex-
pansion. It can also be non-contiguous, accepting
modifiers or some other intervening elements. All
these cases are actually similar to those of the ‘ex-
pansion’ subset except those of the form:
bacterie aerobie/bacterie gram-negatif aerobie
where bacterie remains the head of the term.
The following examples reproduce the general
cases of the ‘expansion’ subset with additional mod-
ifiers:
arteres/anevrysme artere iliaque,
hepatite b/virus hepatite b canard,
encephalite/virus encephalite equine ouest,
sommeil/troubles sommeil extrinseques,
irrigation/liquide irrigation endocanalaire,
maladie/assurance maladie personne agee.
In some of them, adjectival derivation is involved:
cellules/molecule-1 adhesion cellulaire vascu-
laire,
chimie/produits chimiques inorganiques,
dent/implantation dentaire sous-periostee.
Some relations are characteristic of the language
of chemical compounds:
cytochrome c/ubiquinol-cytochrome c reduc-
tase,
diphosphate/uridine diphosphate acide glu-
curonique,
lysine/histone-lysine n-methyltransferase.
The ‘other’ subset also hosted the following mor-
phosyntactic ambiguity:
cilie/cellule ciliee externe
where the words cilie (noun, an invertebrate organ-
ism) and ciliee (inflected form of adjective cilie,
which characterizes a type of cell) are conflated by
lemmatization. This error is mainly due to the fact
that the MeSH is written with unaccented uppercase
letters: the adjective is actually spelled cilié, which
would be unambiguous here.
5 Synthesis
We presented in this paper a human analysis of auto-
matically, lexically-induced term relations that were
not found in the terminology from which the terms
were obtained (the MeSH thesaurus). This lexical
method considers that a term a0 is probably a par-
ent of a term a1 iff all the words of a0 occur in a1 .
This inclusion test is helped by morphological nor-
malization.
Morphological normalization was found to be
useful not only in identifying the already ex-
isting relations (section 3.2), but also for the
‘new’ relations. This confirms previous work by
Jacquemin and Tzoukermann (1999).
The occurrences of syntactic ambiguity
suggest that morphosyntactic tagging could
be useful. The methods specifically de-
signed for detection of syntactic and morpho-
syntactic term variants (Bourigault, 1994;
Jacquemin and Tzoukermann, 1999) might then be
more efficient and less error-prone. We must be
warned however that this may not be an easy task,
since most of the MeSH terms are not syntactically
well-formed (few determiners and prepositions,
inverted heads) and contain rare, technical words
that are likely to be absent from most electronic
lexicons.
Spurious relations may come from several
sources. A few cases are due to abusive morpho-
logical normalization; errors in term names (trans-
lation errors) were also uncovered. We made a dis-
tinction between ‘head’ and ‘expansion’ positions
of the ‘parent’ term in its ‘child’. One would expect
that relations where the parent is in head position
would be correct; however, this is not always true.
The putative head of a term is sometimes not cor-
rectly identified because of specific thesaural con-
structs (the ‘comma’ form) and chemical constructs
(quinone reductases are a kind of reductases) which
display head inversion, and because of enumera-
tions. An additional situation is that of a term whose
actual syntactic head does not entertain an is-a re-
lation with it (the ‘plaster cat’). Furthermore, the
head word may not have a stable meaning: it may
be syntactically ambiguous (cilie), polysemous (in-
vestissement) or underspecified (acne).
The remaining ‘head’ cases reveal specific mod-
eling options, or ‘ontological commitments’, of the
terminology designers: the relations induced might
be considered semantically valid, but were dis-
carded in the MeSH because of overall structuring
choices. These choices cannot be predicted with the
lexical methods used here, and seem to be the most
resistant to attempts at automatic derivation. They
also show that what is correct is not necessarily use-
ful for a given terminology.
The ‘expansion’ cases may be useful to propose
other relations than is-a: we displayed partitive re-
lations, but left to further work a classification of the
remaining ones. The UMLS semantic network rela-
tions (NLM, 2001b) might be a relevant direction to
look into to represent such links.

References
Olivier Bodenreider, Anita Burgun, and Thomas C.
Rindflesch. 2001. Lexically-suggested hy-
ponymic relations among medical terms and their
representation in the UMLS. In URI INIST
CNRS, editor, TIA’2001 Terminologie et Intelli-
gence artificielle, pages 11–21, Nancy.
Didier Bourigault. 1994. Extraction et structura-
tion automatiques de terminologie pour l’aide à
l’acquisition de connaissances à partir de textes.
In Proceedings of the 9 a0 a1 Conference RFIA-
AFCET, pages 1123–1132, Paris, France, Jan-
uary. AFCET.
James J Cimino. 1996. Coding systems in health
care. In Jan H. van Bemmel and Alexa T. Mc-
Cray, editors, Yearbook of Medical Informat-
ics ’95 — The Computer-based Patient Record,
pages 71–85. Schattauer, Stuttgart.
Natalia Grabar and Pierre Zweigenbaum. 2002.
Lexically-based terminology structuring: a fea-
sibility study. In LREC Workshop on Using Se-
mantics for Information Retrieval and Filtering,
pages 73–77, Las Palmas, Canaries, May. ELRA.
Gregory Grefenstette. 1994. Explorations in Auto-
matic Thesaurus Discovery. Natural Language
Processing and Machine Translation. Kluwer
Academic Publishers, London.
Thierry Hamon, Adeline Nazarenko, and Cécile
Gros. 1998. A step towards the detection of
semantic variants of terms in technical docu-
ments. In Christian Boitet, editor, Proceedings
of the 17 a0 a1 COLING, pages 498–504, Montréal,
Canada, 10–14 August.
Marti A. Hearst. 1992. Automatic acquisition of
hyponyms from large text corpora. In Antonio
Zampolli, editor, Proceedings of the 14 a0 a1 COL-
ING, pages 539–545, Nantes, France, 23–28 July.
INSERM, 2000. Thésaurus Biomédical
Français/Anglais. Institut National de la
Santé et de la Recherche Médicale, Paris.
Christian Jacquemin and Didier Bourigault. 2002.
Term extraction and automatic indexing. In Rus-
lan Mitkov, editor, Handbook of Computational
Linguistics. Oxford University Press, Oxford. To
appear.
Christian Jacquemin and Évelyne Tzoukermann.
1999. NLP for term variant extraction: A syn-
ergy of morphology, lexicon, and syntax. In
Tomek Strzalkowski, editor, Natural language
information retrieval, volume 7 of Text, speech
and language technology, chapter 2, pages 25–
74. Kluwer Academic Publishers, Dordrecht &
Boston.
David D. McDonald. 1993. Internal and external
evidence in the identification and semantic cate-
gorization of proper names. In Branimir Bogu-
raev and James Pustejovsky, editors, Corpus Pro-
cessing for Lexical Acquisition, pages 61–76.
MIT Press, Cambridge, MA.
Stuart J Nelson, Douglas Johnston, and Betsy L
Humphreys. 2001. Relationships in medical sub-
ject headings. In Carol A Bean and Rebecca
Green, editors, Relationships in the organization
of knowledge, New York. Kluwer Academic Pub-
lishers.
National Library of Medicine, Bethesda, Mary-
land, 2001a. Medical Subject Headings.
www.nlm.nih.gov/mesh/meshhome.html.
National Library of Medicine, Bethesda, Mary-
land, 2001b. UMLS Knowledge Sources Manual.
www.nlm.nih.gov/research/umls/.
Patrick Séguéla and Nathalie Aussenac. 1999. Ex-
traction de relations sémantiques entre termes
et enrichissement de modèles du domaine. In
Régine Teulier, editor, Actes de IC’99, June.
Pierre Zweigenbaum. 1999. Encoder l’information
médicale : des terminologies aux systèmes de
représentation des connaissances. Innovation
Stratégique en Information de Santé, (2–3):27–
47.
