A Method for Abstracting Newspaper Articles by Using Surface Clues 
Hideo Watanabe 
IBM Research, Tokyo Research Laboratory 
1623-14, Shimotsuruma, Yamato-shi, Kanagawa-ken 242, JAPAN 
watanabe@trl.ibm.co.j p 
Abstract 
This paper describes a system which automatically 
creates an abstract of a newspaper article by selecting 
important sentences of a given text. To determine the 
importance of a sentence, several superficial features 
are considered, and weights for features are determined 
by multiple-regression analysis of a hand processed cor- 
pus. 
1 Introduction 
The rapid expansion of the Internet enables us to 
easily access a lot of information sources in the world. 
The ability to browse information quickly is therefore a 
very important feature of an information retrieval and 
navigation system. Abstraction of a document is one 
useful tool for quick browsing of textual information. 
Generally, an abstract can be considered to be a con- 
cise text giving an outline of the original text. Creat- 
ing an abstract requires deep semantic processing with 
broad knowledge, and the strategy for generating an 
abstract depends on the type of target text. Abstracts 
created by humans tend to differ according to their 
creators' background knowledge and interests. Fur- 
thermore, as stated in \[6\], the same person is likely 
to create different abstracts of the same text at dif- 
ferent times. Simulating this human process is clearly 
outside the area that can be dealt with by current com- 
putational linguistics. There are, however, some cases 
in which an abstract can be created by using surface 
clues to make conjectures as to which portions are the 
most important without using deep semantic process- 
ing. 
The most practical way to create an abstract is thus 
to determine the most important portions by using sur- 
face clues. There are two lines of research based on this 
approach: one analyzes some aspects of a text's struc- 
ture, such as the rhetorical structure \[7\], and selects 
some sentences according to this structure \[5, 3\]; the 
other analyzes surface features for each sentence in a 
given text and selects the most important sentences 
according to some heuristics \[6, 1, 9\]. In methods of 
former type, the rhetorical structure is appropriate for 
a relatively small set of sentences such as a paragraph, 
but it does not give enough information to create an 
abstract for a large set of sentences. In methods of the 
latter type, the validity of the heuristics is uncertain 
when the target text is changed. Therefore, this paper 
proposes a method for selecting important sentences by 
using an equation based on surface features and their 
weights, and a method for determining these weights 
by multiple-regression analysis of abstracts created by 
humans. The target texts of this method are Japanese 
newspaper articles. 
2 Surface Features of a Sen- 
tence 
The proposed method is to create an abstract by 
determining important sentences according to features 
extracted from each sentence. For each sentence in 
a given Japanese newspaper article, the following fea- 
tures 1 are analyzed: 
• Important Keywords: 
An important keyword is defined as a keyword 
that appears in another sentence or in a title. The 
number of points for this feature is tile total num- 
ber of occurrences of impot'tant keywords. 
• Tense: 
The tense of a sentence is analyzed as past or 
present. This feature gives 1 point for present, 
and 0 for past. 
l blest of these features were proposed in the previous studies. 
Keywords were proposed in \[6\], sentence location was proposed 
in \[1\], sentence type was proposed in \[1, 9\], etc., and rhetori- 
ca\] relations were proposed in studies using rhetorical structures 
sud, as \[a\]. 
974 
• Type of a Sentence: 
Sentence types are fact, conjecture, or insistence. 
This feature gives 0 points for fact, 1 for conjec- 
ture, and 2 for insistence. 
• Rhetorical Relation: 
The rhetorical relations to the preceding context 
is analyzed as example, adverse, parallel, compar- 
ison, or connection. This feature gives 1 point for 
reason, 2 for example, and 0 for others. 
• Distance from the beginning of a text: 
In general, sentences located near the beginning of 
a text tend to be important. Therefore, sentences 
in the first paragraph are given 5 points for this 
feature, sentences in the next paragraph 4, and so 
on. 
• Distance from the end of a text: 
Sentences located near the ending of a text also 
tend to be important. Therefore, sentences in the 
last paragraph are given 5 points for this feature, 
sentences in the previous paragraph 4, and so on. 
The tense of a sentence is simply determined to be 
past if it has "ta" (an inflection for the past tense) in 
the last phr~e3 The reason why tense is used is that 
sentences stating about the current fact seem to be 
more important than ones about the past fact in the 
context of editorial articles. 
The sentence type is determined by checking special 
expressions in the last phrase, a For instance, if the fi- 
nal phrase contains "bekida" ("should") or "nakereba- 
naranai" ("must"), then its sentence type is insistence; 
if it contains "darou" ("probably ..."), then its type is 
conjecture; otherwise, its type is fact. Examples of spe- 
cial expressions used to determine sentence type are as 
follows: 
• Conjecture: kamosirenal (may), kanenai (be capa- 
ble of), souda (likely to), youda (likely to), darou 
(probably), etc. 
• Insistence: tai (want to do), hosii (want someone 
to do), bekida (should), nakereba-naranai (must), 
taisetu-dearu (important), hituyouda (necessary), 
etc. 
2 In this method, past does not imply the past tense lit a strict 
sense but rather ;the sentence is not in the present tense. In 
Japanese, "ta" implies the past tense, completion, and so on. 
Most cases are actual instances of the past tense. 
nit is sufficient to check in fire last phrase for Japanese sen- 
tences, because a predicative phrase is always located at the end 
of a Japanese senteltce. Therefore, another strategy is needed 
for languages in which a predicative phrase may be located in 
the middle of a sentence. 
The rhetorical relation is determined by checking 
special expressions both in the first phrase and in the 
last phrase of a sentence. For instance, if "sitakarada" 4 
is found in the last phrase, then the rhetorical relation 
is reason, and if the conjunction "sikasi" ("but") is 
found, then the rhetorical relation is adverse. 5 Exam- 
ples of special expressions used to determine rhetorical 
relations are listed below: 
• Example: tatoeba (for instance), nado (etc.), etc. 
• Adverse: sikasi (but), tokoroga (however), etc. 
• Comparison: koreni-taisi (while), etc. 
• Parallel: mata (further), sarani (in addition), etc. 
• Reason: karada (because), tameda (because), etc. 
3 Process of Creating an Ab- 
stract 
The basic method for creating an abstract in most 
previous studies has been to analyze the sentences of a 
text in terms of some surface features, and a heuristic 
to determine the most important sentences on the basis 
of these features. 
The method proposed in this paper formalizes the 
above approach so that the importance of each sentence 
is calculated as the sum of feature points multiplied by 
their feature weights. The most important sentences 
are then extracted as an abstract. The importance S 
of a sentence is calculated as follows: 
r~ 
i=t 
where a is a constant, P/ is the number of points as- 
signed to the i-th feature, which is normalized to be 
between 0 and 1, and Wi is the weight assigned to the 
i-th feature. 
The steps in creating an abstract are as follows: 
1. For each sentence, calculate the importance. 
2. Select the sentence that has the highest impor- 
tance value among the unselected sentences. 
3. If the selected sentence sl has another sentence 
s2 in the previous context that is related to st by 
any rhetorical structure, then s2 is also selected 
and marked. 
41n English, this expression corresponds to "because" in the 
first phrase. 
5These checking of sentence types and rhetorical relations are 
based on \[10\]. 
975 
4. If the ratio of the number of selected sentences to 
the number of sentences in the text exceeds the 
specified one, then terminate this process; other- 
wise, goto 2. 
These steps select sentences on the basis of their 
importance value, but they also respect the rhetori- 
cal structure to some extent (step 3), because if the 
rhetorical structure is totally ignored, the output text 
will be awkward to read. 
4 A Method for Determining 
the Weights of Features 
Most previous systems can be considered to deter- 
mine the weights of features according to human intu- 
ition. On the other hand, this paper proposes a method 
for determining the wieghts of features by multiple- 
regression analysis of correct examples, which are ab- 
stracts created by testers. A tester selects important 
sentences that should be included in an abstract. The 
importance value of a sentence is defined as the number 
of supporters (testers who selected it as an important 
one) divided by the total number of testers. Let this 
importance value be S; we then get the following equa- 
tion for each sentence: 
S=a+LWI*Pi 
iml 
where, a is a constant, Pi is the number of points as- 
signed to the i-th featnre which is normalized to be 
between 0 to l, and Wi is the weight assigned to the 
i-th feature. 
In this equation, Wi is the only variable. There- 
fore, the feature weight Wi is calculated by multiple- 
regression analysis. 
5 Experiment 
We conducted an experiment to check the validity of 
the proposed method. 
The testers were divided into two groups, A and B, 
each consisting 10 people. Those in group A selected 
important sentences (about 1/3 of the article) in 5 ed- 
itorials and 3 general articles from the Nikkei News- 
paper. Those in group B selected important sentences 
(about 1/3 of the article) in 3 editorials and 3 general 
articles, which were different from those used for group 
A. One of the editorials and one of the general articles 
Feature 
Constant 
Keyword 
Tense 
Type 
Relation 
Location 
(Beginning) 
Location 
(Ending) 
General 
Article 
Weight Weight 
Set 1 Set 2 
0.0 0.183 
1.0 0.216 
0.3 -0.180 
0.3 -0.331 
-1.0 0.127 
1.0 O.437 
1.0 -0.015 
Editorial 
Article 
Weight 
Set 1 
0.0 
1.0 
0.3 
1.0 
-1.0 
1.0 
1.0 
Weight 
Set 2 
0.039 
0.151 
0.046 
0.089 
-0.279 
0.242 
0.214 
Table h Weight Set of Features (General and Editorial 
Articles) 
used for group B are shown in Figures 1 (a) and 2 (a), 
respectively. In each of these figures, the first number 
is a sentence number, the second number is the number 
of supporters in group B, and the last part is a rough 
English translation. G 
Table 1 shows two weight sets; weight set 1 was cre- 
ated by the author in such a way that sentences located 
near the beginning and end are regarded as important, 
sentence importance is not proportional to points for 
rhetorical relation, and the importance of insistence- 
type sentences is higher in editorials than in general 
articles\] Weight set 2, on the other hand, was calcu- 
lated from the results obtained from group A by the 
method described in the previous section. Weight set 
2 for general articles implies that sentences near the 
beginning are more important than ones near the end, 
and insistence-type sentences are less important, and 
so on. On the other hand, weight set 2 for editori- 
als implies that sentences both near the beginning and 
near the end are important, and that insistence-type 
sentences are important, s 
To check the validity of these weight sets, we com- 
pared the abstracts created by the system, using weight 
set 1 and 2, from the articles supplied to group B, with 
6This translation was made by the author, and is not autho- 
rized by Nikkei Newspaper K.K. 
7This weight set 1 corresponds to the way used in previ- 
ous studies, where weights are determined according to human intuition. 
8The weight set calculated in this method can be used as basic 
materiM for creating a practical system, because it is difficult to ask enough people to do this experiment to ensure that the result 
is statistically meaningful. However, the generM tendency can be extracted, and the weight set is determined on the basis of 
this experiment. 
976 
the abstracts created by group B. For the general arti- 
cle in Figure 1 (a), the three most important sentences 
(roughly 1/3 of the article) determined by using the 
weight sets 1 and 2 are listed in Figures 1 (b) and (c), 
respectively. In this case, the three most important 
sentences selected by grou t) B were 0, 2, and 3. Like- 
wise, for the editorial in Figure 2 (a), the eight most 
iml)ortant sentences (roughly 1/a of the article) deter- 
mined by using weight sets l attd 2 are listed in Figures 
2 (b) and (c), respectively, in this case, the eight most 
important sentences selected by grou I) 11 are 0, 2, 3, 12, 
15, 20, 21, 22. Here, we introduce the following metric 
of estrangement to check which abstract is most similar 
to the result of group B: 
Estrangement = ~.,,(the number of supporters of a 
sentence si) - ~_~.</(the number of Sul)porters of a 
sentence s j) 
where s{ is a sentence that is included in an abstract by 
group B but not in an abstract created by the system, 
and s./ is a sentence that is not included in an abstract 
by group B but is included in an abstract created by 
the system. 
The estrangements of the articles in Figures 1 and 2 
are as follows: From this result, the weight set 2 calcu- 
14 
13 
Weight Set 1 Weight Set 2 
6 
12 
to adjust these heuristics for the given text. This pa- 
per proposed a method for this adjustment; that is, 
a method for determining weights of surface features 
by multiple-regression analysis of abstracts created by 
human. By using this method, a system can have an 
ability to be applied to a variety of texts. 
7 Conclusion 
This paper has proposed a method for creating an 
abstract by using surface features and their weights 
to select important sentences, and a method for de- 
termining these \[eature weights by multiple-regression 
analysis of abstracts created by humans. By using the 
proposed method to calculate feature weight, this sys- 
tem can be applied to other types of texts, and gives 
results ntore similar to those of a human process than 
a set of weights based on human intuition. 
This abstract creation system is currently used in an 
informatioll navigation assistance system \[8\] as a tool 
for quickly viewing the contents of newspaper articles. 
References 
\[1\] 
\[2\] 
\[,3\] 
luted by multiple-regression analysis is more similar to 
the human selection than the weight set 1 created ac- 
cording to the author's intuition. For the other general \[4\] 
articles used with group B, the estrangement values of 
weight set 2 are also better than those of weight set \[q 
1. In the other editorials, the estrangement values arc 
comparable. This implies that the weight set I is not 
such a bad estimate for editorials. \[~\] 
6 Discussion \[71 
Edmmtdson, lI. I'., ':New Methods in Automatic Extracting, 
" Journal of the Association for Computing Machiuery, Vol. 
16, No. 2, pp. 264-285, April 1969. 
Kita, S., "A System for Summarization of au Explanatory 
Text" (itl Japanese), Report 63-6 of NLWG of hlformation 
Processing Society of Japan (IPSJ), 1987. 
Ono, K., Sumita, K., and bliike, S., ':Abstract Generation 
based on Rhetorical Structure Extraction," Proc. of Col- 
ing'94, Vo\[.1, pp. 344-348, 1994. 
Tamura, '17., and Tamura, N., "A Summary Generation 
based ou The Form of Text" (in Jap~.aese), Report 92-t 
of NI,WG of IPSJ, 1992. 
So far, most systems for creating an abstract of a 
text has been selected important sentences by some 
heuristics on the basis of surface features. I\[owever, \[8\] 
most of these heuristics were derived from human in- 
tuition, and the validity of them are uncertain if the 
target text is changed. As mentioned in the introduc- \[9\] 
tion, the strategy of an abstraction should be changed 
accordhlg to the given text. Therefore, it is needed 
Schank, 1{. and Abelsoll, A., "Script Plans Goals and Um 
derstandlng," Lawrextce F, rlba.um Associates, IIillside, New 
Jersey, 1977. 
Luhn, H. P., ';The Automatic Creation of Literature AI> 
stracts," IIIM Journal of Research and Development, Vol. 
2, No. 2, pp. 159-165, 1958. 
Mann, W. C., '"Rhetorical Structure Theory: Description 
and Construction of Text Structure," In G. Kempen, edi- 
tor, Natural Lauguage Generation, pp. 279-300, Martiuus 
Nihjhoff Publishers, 1987. 
Morohashi, M., Takedx, K., and et. al., '~\[nformation Outlin- 
ing - Filling tile Gap between Visualization and Navigation 
in Digital Libraries," Proc. of lilt. Symposium o0. Digital 
Libraries 1995, pp. 151-158, 1995. 
Yanmmoto, K., blasuyama, S., and Naitou, S., "GREEN: 
All Experimental System Generating Sutllltlary of Japanese 
Editorials by Combining blultiple Discourse Characteristics" 
(in Japanese), Report 99-3 of NLWG of IPSJ, 1994. 
977 
\[101 Watanabe, H., Tsujii, J., and Nagao, M., "A Method for 
Analyzing Text Structure by Using Surface Clues of Sentence 
" (in Japanese), Proc. of 32nd Convention of IPSJ, pp.1633- 
163-t, 1986. 
Title: ~%~\]~MP U~',' n :,. IBM, ~:~-----~. ?~ 
~tm~:J~o (IBM to release PC equipped with the latest MPU, 
featuring low cost and fast processing) 
0 (10) \[--:~- ~-# 1 0 H =~2g~f6\] ~I BMt~J-H, ~.~'w 4 
poration announced on the 10th of this month that a personal 
computer equipped with the latest "PowerPC" microprocessor 
will be released next summer.) 
(First, a notebook PC will go on sale; this will be followed by 
two types of desktop PC.) 
kN~avl,7o =/>" 1::*=- ~ -~,l~,~n-I~nno (The PowerPC" is used as 
a centrM part of a computer. It is cheap aud has high processing 
power, and is said to be a key to IBM's recovery.) 
Y£o (Since IBM announced its plan to sell personal computers 
equipped PowerPCs, other PC makers in the world are likely to 
take countermeasures.) 
4 (3) ~-~T~---_.~,~IIC D--R om. "~4 ~'. ~ff-PJ'J-~-'4 
(The above three types of PC will provide additional multimedia 
functions by including CD-ROM, microphone, stereo audio, and 
voice recognition functions as standard features.) 
5(0) OS (~g'27 b) liI BM69 \[OS/2J a)ttg,. ~'74 ¢, 
t~','7 bo re\] 4>'Fe2XNTJ, 4\]-:,. v4~m->X~AX'© \['\] 
9 ') xJ %: ~'l.= ~o)bj'l~,-e-~ 7o 3:-5 l:.~Ya o (IBM's OS/2, Microsoft's 
Windows NT, and Sun Microsystems' Solaris will be installed as 
operating systems.) 
6 (O) ;¢~-PCIII BM, 7.~7"Jt,=~ y ~*:~-P, ~b~-9© ~ _ 
Uo ("PowerPC" is a RISC-type MPU developed by IBM, Apple, 
and Motorola.) 
7 (2) e~'\] ~ Y~MP U~ff~T~h©}~.~RI: *; o ~:~'~ 4 >T-)t, 
~MP U~:~}E~To\]tgg¢)~@n~2, \]t~}~ • \]~&}~/~Trj~fSo (It is 
intended to compete with Intel CPUs, which are de-facto stan- 
dards in the PC microprocessor market. Its main advantages are 
low price and fast processing.) 
"~Ta~\]'~'~o (The second largest PC maker, Apple, has 
announced a plan to release a "PowerPC"-based PC next year.) 
9~'oho (The largest PC maker. IBM. has already released a 
PowerPC-based workstation, but has not announced an3' corre- 
sponding plan for PCs.) 
10(1) I BMIi~¢7- P C @9~J-$fdP\]'"O*; <, ¢~\] ¢'7 -~ :, a-)¢~ 
~) ~e~.~g)To~:{'~o (IBM plans to sell PowerPCs to other vendors, 
license the ~ technology, ahd create a family of PowerPC-based 
PC.) 
(Together with Apple and others, IBbl aims to gain at least a 
20% share in the PC market for PowerPC-based PC.) 
(a) Origi~,al Article 
2 (5) t~f~'U~P~N_~©~j~'zerT-PCI$, I BM~©2-'F'~ 
11 (3) 7'~7"~P*3~'~b-~f~ZlT--PC~i~'\]~:,~:~k'~, ~: 
(b) Abstract by Weight Set 1 
o (~o) \[-::~-~--~'10H=~\] )~IBMI~TH, ~J~ 
~o 
(c) Abstract by Weight Set 2 
Fig. h An Example of Abstract of General Article 
(Nikkei Newspaper, 1 Nov. 1993) 
Title: \[{~i~\[{l~:J ~i~F)~ L ~;sl%g)f,: F 4 'Y (~J:~) (Germany de- 
termined to shut out economic refugees.) 
V- 3:7og\]q~A,©~¢k~6 ~;~ C ~ ~:/., L:~ (The German Diet has 
revised the constitution to prohibit immigration for economic 
reasons such as poverty.) 
.5 ~b~J'/',~o (The aim of this move is to shut out economic refugees 
and accept only political refugees.) 
:~di¢3 ~':6~f9~,-(, .~¢2,~:~1".¢b~,4,_3:-50 (This retreat from 
idealism is disappointing, but in view of the current situation of 
Germany, it is an inevitable measure.) 
©f~o (The restriction, which will come into effect in July, will 
prohibit refugees from "countries without persecution" (e.g. Ro- 
mania, Bulgaria, and Hungary) from being granted entry except 
in special cases, and will repatriate political refugees through 
"safe countries" (e.g. western European countries, Poland, and 
Czechoslovakia) which permit pohtical refugees.) 
4 (i) ~.9 *~ ~'~J~l I:~ b t:~ ~ ~, c~ ~C~ 
~7o~'{'~{<¢t,,A~, "5~'~7oo (The logic behind this is that 
there cannot be political refugees from countries that have been 
converted into democratic nations by the East European Revo- 
lution, and so on.) 
l~@.*~:-'~;tqfzof:.o (Clause 2, Article 16 of the Basic Law es- 
tablished after WWlI, in 1949, was generous to refugees stating 
that people persecuted politically had a right to be protected.) 
was based on the reflection that anti-foreign policies in the 'Nazi' 
era had hurt foreign nations and produced many refugees from 
Germany.) 
< o f.: ~ ~o I,~.5 ~ (It is also said that this clause was created out 
of strong consideration for socialist states.} 
f.:o (However, Germany's situation has been totally changed by 
the unification of Germany and the end of the cold war.) 
(The number of refugee applicants arriving in major Western 
European countries last year reached 700,000, among whom the 
978 
number arriving in Germany, which has loose restrictions, was 
440,000, over 60% of the total for Europe.) 
10 (0) U.a)m\]:JI21$Z\]7:\]~.-T-)k.7)¢V 4 'Y}21~.@"~l~a~b, \[~--+\]< 
people applied to Germany for refugee status this April; of these, 
410,000 were interviewed but only 700 were accepted.) 
~gE~7)'g69Xa'~d~&G~l-Zoo (73% of applicants are from Eastern 
Europe or ex-soviet-bloc countries such as Romania, and former 
Yugoslavia.) 
12 (0) --%Z*, >\[<i~.i.~\[ ~a ~:~d~/-.: %: ~ 'o (at the same tinle, titere 
is a constant flow of entrants.) 
13 (5) g 4 "9lJ:~'£, i~l{f~.~.~.& g~,~7~74<iS{,TIza~7~o (Germany 
is now said to be suffering the worst recession since WWII.) 
14 (0) IH~F 4 "~f.~a)9£~!{~11~\[~lH -c-{~ • --%/27) ¢, {(}:i/\]~\]\[~.£7)¢~ 
~cv,7olH~/{ F -1' '2~.~-~li--~ • g%~'o (Tim unemployment 
rate in April was 7.1% in former West Germany, and 14.7% in 
former East Germany.) 
~, ~ -} 74<~7)~ ~ 7o 7), #o/'.50 (There is a growing anti-forelgn tendency 
manifested in attacks on foreigners by ultra-rightist groups, re- 
suiting from anxiety that masses of immigrants will take native 
people's jobs.) 
(Refugees enter government-provided accommodation and live 
there until ttleir inverviews are completed, with their expenses 
borne by states and cities.) 
(For this reason, many regional governments have appealed for 
the numbers of refugees to be restricted.) 
18 (0 u ) i~l,~_g, a)DJ~i~J Ca~.~f 3:5 ~ t,, ~ ~ag~£©i~>-l=ll, 
,~.-~g.~o~aT),, ~%. ¢i~A~a)~a)~ < g ~2 bf:o (Most 
Diet members in coalition government parties and social liberal- 
ist party approved the revision of rite Basic Law, which intposes 
restrictions on refugees similar to those of other major Western 
European countries.) 
Z'U'L"0©~/:~7)~.:~ok~ b'")t~l\]~l~.¢'~fTao (Tile move is also 
iutended to be a post-war process and the settlemeut after the 
Cold War, aud gives tire impression of the end of an era in Ger- 
many.) 
-9"("u"7o o (Other major Western European countries such as 
France have also decided to impose more restrictions on eco- 
nonfic refugees.) 
21 (4) t~tt,,l$1©A4 r) {-a T) , &- I~J -"-.~ ~:a -) & -¢.5 ¢)lat g ~Sa)~ b {?~ 
?.9)< 7k~a)~$!~jlat?~aL~bo (It is natural for poor people 
to try to go to rich countries, but the movenmnt of many people 
produces confusion and friction.) 
tant to Mleviate the conditions that produce econonfic refugees 
through world-wide cooperation, and it may be inevitable for 
European countries to impose some level of restricts.) 
g)-C~7oa)t~,~7){a~7~ ")o (However, it is not appropriate to 
apply these examples to Japan, because Japan's circumstances 
are different from those in Europe.) 
(a) OrigiuM Article 
u,Sa~ft~o 
2 (8) ~_~,$~a)i{i~_Iz~tZT){, A~,II4~*~_~, V 4 "77k.#,~)10~ 
a (6) -b~17),gZ~~cTo~lJ{~l£, Fi~a)gu,\[~\] (;b-7--7, 
f~o 
"CW6o 
<'~ b, #A#l~7)¢--,~e)~llA,$'~.lf,5© g.,eb,k{~.:~ v,o 
(b) Abstract by Weight Set 1 
3 (6) -{5~JT)"6~)~iTTa~bj~lJ.{~'n~{~, \[~=m©exv,V,l\] (m--~'m7, 
(a) Abstract by Weight Set 2 
Fig. 2: An Example of Abstract of Editorial 
(Nikkei Newspaper, 1 Jun. 1993) 
979 
