A Class-oriented Approach to Building a Paraphrase Corpus
Atsushi Fujita
Graduate School of Informatics,
Kyoto University
fujita@pine.kuee.kyoto-u.ac.jp
Kentaro Inui
Graduate School of Information Science,
Nara Institute of Science and Technology
inui@is.naist.jp
Abstract
Towards deep analysis of composi-
tional classes of paraphrases, we have
examined a class-oriented framework
for collecting paraphrase examples, in
which sentential paraphrases are col-
lected for each paraphrase class sep-
arately by means of automatic can-
didate generation and manual judge-
ment. Our preliminary experiments on
building a paraphrase corpus have so
far been producing promising results,
which we have evaluated according to
cost-efficiency, exhaustiveness, and re-
liability.
1 Introduction
Paraphrases are alternative ways of conveying the
same content. The technology for paraphrase
generation and recognition has drawn the atten-
tion of an increasing number of researchers be-
cause of its potential contribution toabroadrange
of natural language applications.
Paraphrases can be viewed as monolingual
translations. From this viewpoint, research on
paraphrasing has adapted techniques fostered in
the literature of machine translation (MT), such
as transformation algorithms (Lavoie et al., 2000;
Takahashi et al., 2001), corpus-based techniques
for paraphrase pattern acquisition (Barzilay and
McKeown, 2001; Shinyama and Sekine, 2003;
Quirk et al., 2004), and fluency measurements
(Lapata, 2001; Fujita et al., 2004).
One thing the paraphrasing community is still
lacking is shared collections of paraphrase ex-
amples that could be used to analyze problems
underlying the tasks and to evaluate the perfor-
mance of systemsunder development. Toour best
knowledge, the paraphrase corpus developed by
Dolan et al. (2004) is one of the very few collec-
tions available for free1. Development of para-
phrase corpora raises several issues: what sorts
of paraphrases should be collected, where para-
phrase examples can be obtained from, how the
coverageandqualityof thecorpuscanbeensured,
how manual annotation cost can be effectively re-
duced, and how collected examples should be or-
ganized and annotated.
Obviously these issues should be discussed
with the purpose of each individual corpus taken
into account. In this paper, we address the is-
sues of building a gold-standard corpus that is to
be used to evaluate paraphrase generation models
and report on our preliminary experiences taking
Japanese as a target language. Our approach is
characterized by the following:
 We define a set of paraphrase classes based
on the syntactic features of transformation
patterns.
 We separately collect paraphrase examples
for each paraphrase class that are considered
to be linguistically explainable.
 We use a paraphrase generation system to
exhaustively collect candidate paraphrases
from a given text collection, which are then
manually labeled.
15801 sentence pairs from their comparable corpus have
been judged manually and available from
http://research.microsoft.com/research/nlp/msr paraphrase.htm
25
2 Goal
Paraphrases exhibit a wide variety of patterns
ranging from lexical paraphrases to syntactic
transformations and their combinations. Some
of them are highly inferential or idiomatic and
do not seem easy to generate only with syntactic
and semantic knowledge. Such groups of para-
phrasesrequireustopursuecorpus-basedacquisi-
tion methods such asthosedescribedin Section 3.
More importantly, however, we can also find
quite a few patterns of paraphrases that exhibit a
degree of regularity. Those groups of paraphrases
have a potential to be compositionally explained
by combining syntactic and semantic properties
of their constituent words. For instance, the fol-
lowingparaphrases2 inJapaneseareconsideredto
be of these groups.
(1) s. eiga-ni shigeki-o uke-ta.
film-DAT inspiration-ACC to receive-PAST
I received an inspiration from the film.
t. eiga-ni shigeki-s-are-ta.
film-DAT to inspire-PASS-PAST
I was inspired by the film.
(2) s. sentakumono-ga soyokaze-ni yureru.
laundry-NOM breeze-DAT to sway-PRES
The laundry sways in the breeze.
t. soyokaze-ga sentakumono-o yurasu.
breeze-NOM laundry-ACC to sway-PRES
The breeze makes the laundry sways.
(3) s. glass-ni mizu-o mitashi-ta.
glass-DAT water-ACC to fill-PAST
I filled water into the glass.
t. glass-o mizu-de mitashi-ta.
glass-ACC water-IMP to fill-PAST
I filled the glass with water.
(4) s. kare-wa kikai-sousa-ga jouzu-da.
he-TOP machine operation-NOM be good-PRES
He is good at machine operation.
t. kare-wa kikai-o jouzu-ni sousa-suru.
he-TOP machine-ACC well-ADV to operate-PRES
He operates machines well.
(5) s. heya-wa mou atatamat-teiru.
room-TOP already to be warmed-PERF
The room has already been warmed up.
t. heya-wa mou atatakai.
room-TOP already be warm-PRES
The room is warm.
2For each example, “s” and “t” denote an original sen-
tence and its paraphrase, respectively.
In example (1), a verb phrase, “shigeki-o uke-
ta (to receive an inspiration),” is paraphrased into
a verbalized form of the noun, “shigeki-s-are-ta
(to be inspired).” We can find a number of para-
phrases that exhibit a similar pattern of syntactic
transformation in the same language and group
such paraphrases into a single class, which is
possibly labeled “paraphrasing of light-verb con-
struction.” Likewise, paraphrases exemplified by
(2) constitute another class, so-called transitiv-
ity alternation. Example (3) is of the locative
alternation class and example (4) the compound
noun decomposition class. In example (5), a verb
“atatamaru (to be warmed)” is paraphrased into
its adjective form, “atatakai (be warm).” Para-
phrases involving such a lexical derivation are
also in our concern.
One can learn the existence of such groups
of paraphrases and the regularity each group ex-
hibits from the linguistic literature (Mel’ˇcuk and
Polgu`ere, 1987; Jackendoff, 1990; Kageyama,
2001). According to Jackendoff and Kageyama,
for instance, both transitivityalternation and loca-
tive alternation can be explained in terms of the
syntactic and semantic properties of the verb in-
volved, which are represented by what they call
Lexical Conceptual Structure. The systematicity
underlying such linguistic accounts is intriguing
also from the engineering point of view asit could
enable us to take a more theoretically motivated
but still practical approach to paraphrase genera-
tion.
Aiming at this goal leads us to consider build-
ing a paraphrase corpus which enables us to eval-
uate paraphrase generation systems and conduct
error analysis for each paraphrase class sepa-
rately. Our paraphrase corpus should therefore be
organized according to paraphrase classes. More
specifically, weconsider aparaphrasecorpussuch
that:
 The corpus consists of a set of subcorpora.
 Each subcorpus is a collection of paraphrase
sentence pairs of a paraphrase class.
 Paraphrases collected in a subcorpus suffi-
ciently reflect the distribution of the occur-
rences in the real world.
Given a paraphrase class and a text collection,
the goal of building a paraphrase corpus is to col-
lect paraphrase examples belonging to the class
26
as exhaustively as possible from the text collec-
tion at a minimal human labor cost. The resultant
corpus should also be reliable.
3 Related work
Previous work on building paraphrase corpus
(collecting paraphraseexamples) canbeclassified
into two directions: manual production of para-
phrases and automatic paraphrase acquisition.
3.1 Manual production of paraphrases
Manual production of paraphrase examples has
been carried out in MT studies.
For example, Shirai et al. (2001) and
Kinjo et al. (2003) use collections of Japanese-
English translation sentence pairs. Given
translation pairs, annotators are asked to produce
new translations for each side of the languages.
Sentences that have an identical translation
are collected as equivalents, i.e., paraphrases.
Shimohata (2004), on the other hand, takes a
simpler approach in which he asks annotators to
produce paraphrases of a given set of English
sentences.
Obviously, if we simply asked human annota-
tors to produce paraphrases of a given set of sen-
tences, the labor cost would be expensive while
the coverage not guaranteed. Previous work,
however, has averted their eyes from evaluating
the cost-efficiency of the method and the cover-
age of the collected paraphrases supposedly be-
cause their primary concern was to enhance MT
systems.
3.2 Automatic paraphrase acquisition
Recently, paraphrase examples have been auto-
matically collected as a source of acquiring para-
phrase knowledge, such as pairs of synonymous
phrases and syntactic transformation templates.
Some studies exploit topically related articles
derived from multiple news sources (Barzilay and
Lee, 2003; Shinyama and Sekine, 2003; Quirk et
al., 2004; Dolan et al., 2004). Sentence pairs that
are likely to be paraphrases are automatically col-
lected from the parallel or comparable corpora,
using such clues as overlaps of content words and
named entities, syntactic similarity, and reference
description, such as date of the article and posi-
tions of sentences in the articles.
Automatic acquisition from parallel or compa-
rable corpora, possibly in combination with man-
ual correction, could be more cost-efficient than
manual production. However, it would not ensure
coverage and quality, because sentence pairingal-
gorithms virtually limit the range of obtainable
paraphrases and products tend to be noisy.
Nevertheless, automatic methods are useful to
discover a variety of paraphrases that need further
exploration. We hope that our approach to corpus
construction, which we present below, will work
complementary to those directions of research.
4 Proposed method
Recall that we require a corpus that reflects the
distribution of the occurrences of potential para-
phrases of each class because we aim to use it for
linguistic analysis and quantitative evaluation of
paraphrase generation models.
Since the issues weaddress here are highly em-
pirical, we need to empirically examine a range
of possible methods to gain useful methodologi-
cal insights. As an initial attempt, we have so far
examined asimplemethod which falls in the mid-
dle of the aforementioned two approaches. The
method makes use of an existing paraphrase gen-
eration system to reduce human labor cost as well
as to ensure coverage and quality:
Step 1. For a given paraphrase class, develop a
set of morpho-syntactic paraphrasing pat-
terns and lexical resources.
Step 2. Apply the patterns to a given text collec-
tion using the paraphrasing system to gener-
ate a set of candidate paraphrases.
Step 3. Annotate each candidate paraphrase with
information of the appropriateness accord-
ing to a set of judgement criteria.
Weusemorpho-syntacticparaphrasingpatterns
derived from paraphrase samples in an analogous
way to previous methods such as (Dras, 1999).
For instance, from example (1), we derive a para-
phrasing pattern for paraphrasing of light-verb
constructions:
(6) s. N-o(⇒V ) V
N-ACC V
t. V (N)
V (N)
whereN is avariable which matcheswith a noun,
V a verb, V (N) denotes the verbalized form of
27
(e) confirmed (revised)
paraphrase
( ) fir  (r i )
r r
(c) annotator’s judge
(correct / incorrect)
( ) t t r’  j
( rr t / i rr t)
(d) error tags( ) rr r t
(a) source sentence( ) r  t
(b) automatically
generated
paraphrase
( ) t ti ll
r t
r r
(c) second opinion
(correct / incorrect)
( )  i i
( rr t / i rr t)
Given
Obligatory
Obligatory
Optional
(f) free comments(f) fr  t
Optional
Figure 1: Annotation schema.
N, and the subscriptedarrowin (6s) indicates that
N-o depends on V.
To exhaustively collect paraphrase examples
from a given text collection, we should not exces-
sively constrain paraphrasing patterns. To avoid
overly generating anomalies, on the other hand,
we make use of several lexical resources. For in-
stance, pairs of a deverbal noun and its transitive
form are used to constrainN andV (N) in pattern
(6). This way, we combine syntactic transforma-
tion patterns with lexical constraints to specify a
paraphrase class. This approach is practical given
the recent advances of shallow parsers.
ForthejudgementonappropriatenessinStep3,
we create a set of criteria separately for each para-
phrase class. When the paraphrase class in focus
is specified, the range of potential errors in candi-
date generation tends to be predictable. We there-
fore specify judgement criteria in terms of a ty-
pology of potential errors (Fujita and Inui, 2003);
namely, we provide annotators with a set of con-
ditions for ruling out inappropriate paraphrases.
Annotators judge each candidate paraphrase
with a view of an RDB-based annotation tool
(Figure 1). Given (a) a source sentence and
(b) an automatically generated candidate para-
phrase, human annotators are asked to (c) judge
the appropriateness of it and, if it is inappropri-
ate, they are also asked to (d) classify the un-
derlying errors into a predefined taxonomy, and
make (e) appropriate revisions (if possible) and
(f) format-free comments.
5 Preliminary trials
To examine how the proposed method actually
work regarding the issues, we conducted prelim-
inary trials, taking two classes of Japanese para-
phrases: paraphrasing of light-verb constructions
and transitivity alternation. This section de-
scribes the settings for each paraphrase class.
We sampled a collection of source sentences
from one year worth of newspaper articles: Ni-
hon Keizai Shinbun3, 2000, where the average
sentence length was 25.3 words. The reason
why we selected newspaper articles as a sample
source was that most of the publicly available
shallow parsers for Japanese were trained on a
tree-bank sampled from newspaper articles, and a
newspaper corpus was available in a considerably
large scale. We used for candidate generation the
morphologicalanalyzer ChaSen4,thedependency
structure analyzer CaboCha5, and the paraphrase
generation system KURA6.
Two native speakers of Japanese, adults grad-
uated from university, were employed as annota-
tors. The process of judging each candidate para-
phrase is illustrated in Figure 2. The first annota-
tor was asked to make judgements on each candi-
date paraphrase. The second annotator inspected
all the candidates judged correct by the first an-
3http://sub.nikkeish.co.jp/gengo/zenbun.htm
4http://chasen.naist.jp/
5http://chasen.org/˜taku/software/cabocha/
6http://cl.naist.jp/kura/doc/
28
Candidate
paraphrase
i t
r r
Correct
Incorrect
1st annotator 2nd annotator
Correct
Incorrect
Correct
Incorrect
Correct
Deferred
Incorrect
Discussion
Unseen
Deferredf
Correctt
IncorrectI t
Label
Figure 2: Judgement procedure.
notator. To reduce the labor cost, only a small
subset of candidates that the first annotator judged
incorrect were checked by the second annotator,
leaving the rest labeled incorrect. Once in sev-
eral days,theannotatorsdiscussedcasesonwhich
they disagreed, and if possible revised the anno-
tation criteria. When the discussion did not reach
a consensus, the judgement was deferred.
5.1 Paraphrasing of light-verb constructions
(LVC)
An example of this class is given in (1). A light-
verb construction consists of a deverbal noun
(“shigeki (inspiration)” in example (1)) governed
by a light-verb (“ukeru (to receive)”). A para-
phrase of this class is a pair of a light-verb con-
struction and its unmarked form, which consists
of the verbalized form of the deverbal noun where
the light-verb is removed.
Let N, V be a deverbal noun and a verb, and
V (N) be the verbalized form of N. Paraphrases
of this class can be represented by the following
paraphrasing pattern:
(7) s. N-{ga, o, ni}(⇒V ) V
N-{NOM, ACC, DAT} V
t. V (N)
V (N)
In the experiment, we used three more patterns to
gain the coverage.
We then extracted 20,155 pairs of deverbal
noun and its verbalized form (e.g. “shigeki (in-
spiration)” and “shigeki-suru (to inspire)”) from
the Japanese word dictionary, IPADIC (version
2.6.3)3. This set was used as a restriction on
nouns that can match with N in a paraphrasing
pattern. On the other hand, we made no restric-
tion on V, because we had no exhaustive list
of light-verbs. The patterns were automatically
compiled into pairs of dependency trees with
uninstantiated components, and were applied to
source sentences with the paraphrase generation
system, which carried out dependency structure-
based pattern matching. 2,566 candidate para-
phrases were generated from 10,000 source sen-
tences.
In the judgement phase, the annotators were
also asked to revise erroneous candidates if pos-
sible. The following revision operations were al-
lowed for LVC:
 Change of conjugations
 Change of case markers
 Insert adverbs
 Append verbal suffixes, such as voice, as-
pect, or mood devices
When pattern (7) is applied to sentence (1s), for
instance, weneedtoaddavoicedevice, “are(pas-
sive),” to correctly produce (1t). In example (8),
on the other hand, an aspectual device, “dasu (in-
choative),” is appended, and a case marker, “no
(GEN),” is replaced with “o (ACC).”
(8) s. concert-no ticket-no hanbai-o hajime-ta.
concert-GEN ticket-GEN sale-ACC to start-PAST
We started to sale tickets for concerts.
t. concert-no ticket-o hanbai-shi-dashi-ta.
concert-GEN ticket-ACC to sell-INCHOATIVE-PAST
We started selling tickets for concerts.
So far, 1,114 candidates have been judged7 with
agreements on 1,067 candidates, and 591 para-
phrase examples have been collected.
5.2 Transitivity alternation (TransAlt)
This class of paraphrases requires a collection of
pairs of intransitive and transitive verbs, such as
“yureru (to sway)” and “yurasu (to sway)” in ex-
ample (2). Since there was no available resource
of such knowledge, we newly created a mini-
mal set of intransitive-transitive pairs that were
required to cover all the verbs appearing in the
source sentence set (25,000 sentences). We first
retrieved all the verbs from the source sentences
using a set of extraction patterns implemented in
the same manner as paraphrasing patterns. Ex-
ample (9) is one of the patterns used, where Nx
matches with a noun, and V a verb.
7983 candidates for the first 4,500 sentences were fully
judged, and 131 candidates were randomly sampled from
the remaining portion.
29
(9) s. N1-ga(⇒V ) N2-ni(⇒V ) V
N1-NOM N2-DAT V
t. no change.
We then manually examined the transitivity of
each of 800 verbs that matched with V, and col-
lected212 pairsof intransitiveverbvi and itstran-
sitive form vt. Using them as constraints, we im-
plemented eight paraphrasing patterns as in (10).
(10) s. N1-ga(⇒Vi) N2-ni(⇒Vi) Vi
N1-NOM N2-DAT Vi
t. N2-ga(⇒Vt(Vi)) N1-o(⇒Vt(Vi)) Vt(Vi)
N2-NOM N1-ACC Vt(Vi)
where Vi and Vt(Vi) are variables that match with
vi and vt, respectively. By applying the patterns
to the same set of source sentences, we obtained
985 candidate paraphrases.
We created a set of criteria for judging ap-
propriateness (an example will be given in
Section 6.4) andrevisionexamplesfor the follow-
ing operations allowed for this trial:
 Change of conjugations
 Change of case markers
 Change of voices
964 candidates have gained an agreement, and
484 paraphrase examples have been collected.
6 Results and discussion
Table 1 gives some statistics of the resultant para-
phrase corpora. Figures 3 and 4 show the number
of candidate paraphrases, where the horizontal
axes denote the total working hours of two anno-
tators, and the vertical axes the number of candi-
dateparaphrases. Thenumbers ofjudged, correct,
incorrect, and deferred candidates are shown.
6.1 Efficiency
2,031 candidate paraphrases have so far been
judged in total and 1,075 paraphrase examples
have been collected in 287.5 hours. The judge-
ment was performed at a constant pace: 7.1 can-
didates (3.7 examples) in one hour. It is hard to
compare these results with other work because
no previous study quantitatively evaluate the effi-
ciency in terms of manual annotation cost. How-
ever, we feel that the results have so far been sat-
isfiable.
For each candidate paraphrase judged incor-
rect, the annotators were asked to classify the un-
derlying errors into the fixed error types ((d) in
Table 1: Statistics of the resultant corpora.
Paraphrase class LVC TransAlt
# of source sentences 10,000 25,000
# of patterns 4 8
Type of lexical resources 〈n, vn〉 〈vi, vt〉
Size of lexical resource 20,155 212
# of candidates 2,566 985
# of judged candidates 1,067 964
# of incorrect candidates 520 503
# of correct candidates 547 461
# of paraphrase examples 591 484
Working hours 118 169.5
Figure 1). This error classification consumed ex-
tra time because it required linguistic expertise
which the annotators were not familiar with.
TransAlt was 1.75 times more time-consuming
than LVC because the definition of TransAlt in-
volved several delicate issues, which made the
judgement process complicated. We return to this
issue in Section 6.4.
6.2 Exhaustiveness
To estimate how exhaustively the proposed
method collected paraphrase examples, we ran-
domly sampled 750 sentences from the 4,500
sentences that were used in the trial for LVC,
and manually checked whether the LVC para-
phrasing could apply to each of them. As a re-
sult, 206 examples were obtained, 158 of which
were those already collected by the proposed
method. Thus, the estimated exhaustiveness was
77% (158/206). Our manual investigation into
the missed examples has revealed that 47 misses
could have been automatically generated by en-
hancing paraphrasing patterns and dictionaries,
while only one example was missed due to an er-
ror in shallow parsing. 34 cases of the 48 misses
could have been collected by adding a couple of
paraphrasing patterns. For example, pattern (11)
verbalizes anounfollowedby anominalizing suf-
fix, “ka (-ize),” as in (12).
(11) s. N-ka-{ga, o, ni}(⇒V ) V
N-ize-{NOM, ACC, DAT} V
t. V (N-ka)
V (N-ize)
(12) s. kore-wa kin’yu-shijo-no kassei-ka-ni
this-TOP financial market-GEN activation-DAT
muke-ta kisei-kanwa-saku-da.
to address-PAST deregulation plan-COP
This is a deregulation plan aiming at the
activation of financial market.
30
 0
 200
 400
 600
 800
 1000
 1200
 0  20  40  60  80  100  120
# of judged candidates
working hours
Judged
Correct
Incorrect
Deferred
Figure 3: # of judged candidates (LVC).
 0
 200
 400
 600
 800
 1000
 0  20  40  60  80  100  120  140  160  180
# of judged candidates
working hours
Judged
Correct
Incorrect
Deferred
Figure 4: # of judged candidates (TransAlt).
t. kore-wa kin’yu-shijo-o
this-TOP financial market-ACC
kassei-ka-suru kisei-kanwa-saku-da.
to activate-PRES deregulation plan-COP
This is a deregulation plan which activates
financial market.
We cannot know if we have adequate para-
phrasing patterns and resources before trials.
Therefore, manual examination isnecessary tore-
finethem tobridgegap between therangeof para-
phrases that can be automatically generated and
those of the specific class we consider.
6.3 Reliability
Ideally, more annotators should be employed to
ensure the reliability of the products, which, how-
ever, leads to a matter of balancing the trade-off.
Instead, we specified the detailed judgement cri-
teria for each paraphrase class, and asked the an-
notators to reconsider marginal cases several days
later and to make a discussion when judgements
disagreed. The agreement ratio for correct candi-
dates between two annotators increased as they
became used to the task. In the trial for LVC,
for example, the agreement ratio for each day
changed from 74% (day 3) to 77% (day 6), 88%
(day 9), and 93% (day 11). This indicates that the
judgement criteria were effectively refined based
on the feedback from inter-annotator discussions
on marginal and disagreed cases. To evaluate the
reliability of our judgement procedure more pre-
cisely, we are planing to employ the third annota-
tor who will be asked to judge all the cases inde-
pendently of the others.
6.4 How we define paraphrase classes
One of the motivations behind our class-basedap-
proach is an expectation that specifying the target
classes of paraphrases would simplify the awk-
ward problem of defining the boundary between
paraphrasesan non-paraphrases. Our trialsfor the
two paraphrase classes, however, have revealed
that it can still be difficult to create a clear cri-
terion for judgement even when the paraphrase
class in focus is specified.
As one of the criteria for TransAlt, we tested
the agentivity of the nominative case of intransi-
tive verbs. The test used an adverb, “muzukara
(by itself),” and classified a candidate paraphrase
as incorrect if the adverb could be inserted im-
mediately before the intransitive verb. For ex-
ample, we considered example (13) as a correct
paraphrase of the TransAlt class whereas (14) in-
correct because the agentivity exhibited by (14s)
did not remain in (14t).
(13) s. kare-ga soup-o atatame-ta.
he-NOM soup-ACC to warm up-PAST
He warmed the soup up.
t. soup-ga atatamat-ta. (correct)
soup-NOM to be warmed up-PAST
The soup was warmed up (by somebody).
(14) s. kare-ga koori-o tokashi-ta.
he-NOM ice-ACC to melt (vt)-PAST
He melted the ice.
t. koori-ga toke-ta. (incorrect)
ice-NOM to melt (vi)-PAST
The ice melted (by itself).
However, one might regard both paraphrases
incorrect because the information given by the
nominative argument of the source sentence is
31
dropped in the target in both cases. Thus, the
problem still remains. Nevertheless, our approach
will provide us with a considerable amounts of
concrete data, which we hope will lead us to bet-
ter understanding of the issue.
7 Conclusion
Towardsdeepanalysisofcompositionalclassesof
paraphrases, we have examined a class-oriented
framework for collecting paraphrase examples,
in which sentential paraphrases are collected for
each paraphrase class separately by means of au-
tomatic candidate generation and manual judge-
ment. Our preliminary experiments on building
a paraphrase corpus have so far been producing
promising results, which we have evaluated ac-
cording to cost-efficiency, exhaustiveness, and re-
liability. The resultant corpus and resources will
be available for free shortly. Our next step is di-
rected to targeting a wider range of paraphrase
classes.
References
Regina Barzilay and Kathleen R. McKeown. 2001.
Extracting paraphrases from a parallel corpus. In
Proceedings of the 39th Annual Meeting of the
Association for Computational Linguistics (ACL),
pages 50–57.
Regina Barzilay and Lillian Lee. 2003. Learn-
ing to paraphrase: an unsupervised approach us-
ing multiple-sequence alignment. In Proceedings
of the 2003 Human Language Technology Confer-
enceandtheNorthAmericanChapteroftheAssoci-
ation for Computational Linguistics (HLT-NAACL),
pages 16–23.
Bill Dolan, Chris Quirk, and Chris Brockett. 2004.
Unsupervised construction of large paraphrase cor-
pora: exploiting massively parallel news sources.
In Proceedings of the 20th International Con-
ference on Computational Linguistics (COLING),
pages 350–356.
MarkDras. 1999. Tree adjoining grammar andthere-
luctant paraphrasing of text. Ph.D. thesis, Division
of Information and Communication Science, Mac-
quarie University.
Atsushi Fujita and Kentaro Inui. 2003. Explor-
ing transfer errors in lexical and structural para-
phrasing. IPSJ Journal, 44(11):2826–2838. (in
Japanese).
Atsushi Fujita, Kentaro Inui, and Yuji Matsumoto.
2004. Detection of incorrect case assignments in
automatically generated paraphrases of Japanese
sentences. In Proceedings of the 1st International
Joint Conference on Natural Language Processing
(IJCNLP), pages 14–21.
Ray Jackendoff. 1990. Semantic structures. The MIT
Press.
Taro Kageyama, editor. 2001. Semantics and syntax
of verb: comparable study between Japanese and
English. Taishukan Shoten. (in Japanese).
Yumiko Kinjo, KunioAono, KeishiYasuda, Toshiyuki
Takezawa, and Genichiro Kikui. 2003. Collec-
tion of Japanese paraphrases of basic expressions
on travel conversation. In Proceedings of the 9th
Annual Meeting of the Association for Natural Lan-
guage Processing, pages 101–104. (in Japanese).
Maria Lapata. 2001. A corpus-based account of reg-
ular polysemy: the case of context-sensitive ad-
jectives. In Proceedings of the 2nd Meeting of
the North American Chapter of the Association for
Computational Linguistics (NAACL), pages 63–70.
Benoit Lavoie, Richard Kittredge, Tanya Korelsky,
and Owen Rambow. 2000. A framework for MT
and multilingual NLG systems based on uniform
lexico-structural processing. In Proceedings of the
6th Applied Natural Language Processing Confer-
ence and the 1st Meeting of the North American
Chapter of the Association for Computational Lin-
guistics (ANLP-NAACL), pages 60–67.
Igor Mel’ˇcuk and Alain Polgu`ere. 1987. A formal
lexicon in meaning-text theory (or how to do lex-
ica with words). Computational Linguistics, 13(3-
4):261–275.
Chris Quirk, Chris Brockett, and William Dolan.
2004. Monolingual machine translation for para-
phrase generation. In Proceedings of the 2004 Con-
ference on Empirical Methods in Natural Language
Processing (EMNLP), pages 142–149.
Mitsuo Shimohata. 2004. Acquiring paraphrases
from corpora and its application to machine trans-
lation. Ph.D. thesis, Graduate School of Informa-
tion Science, Nara Institute of Science and Tech-
nology.
Yusuke Shinyama and Satoshi Sekine. 2003. Para-
phrase acquisition for information extraction. In
Proceedings of the 2nd International Workshop on
Paraphrasing: Paraphrase Acquisition and Appli-
cations (IWP), pages 65–71.
Satoshi Shirai, Kazuhide Yamamoto, and Francis
Bond. 2001. Japanese-English paraphrase corpus.
In Proceedings of the 6th Natural Language Pro-
cessing Pacific Rim Symposium (NLPRS) Workshop
on Language Resources in Asia, pages 23–30.
Tetsuro Takahashi, Tomoya Iwakura, Ryu Iida, At-
sushi Fujita, and Kentaro Inui. 2001. KURA:
a transfer-based lexico-structural paraphrasing en-
gine. In Proceedings of the 6th Natural Language
Processing Pacific Rim Symposium (NLPRS) Work-
shoponAutomaticParaphrasing: Theories andAp-
plications, pages 37–46.
32
