Enhancement of a Chinese Discourse Marker Tagger with C4.5 
Benjamin K. T'sou l, Torn B. Y. Lai 2, Samuel W. K. Chan 3, Weijun Gao 4, Xuegang Zhan 5 
23Languag e Information Sciences Research Centre 
City University of Hong Kong 
Tat Chee Avenue, Kowloon 
Hong Kong SAR, China 
Northeastern University, China 
{ ~rlbtsou, 2ettomlai} @uxmail.cityu.edu.hk, 3swkchan@cs.cityu.edu.hk, 
4wj gao@mail.neu.edu.cn, Szxg@ics.cs.neu.edu.cn 
Abstract 
Discourse markers are complex 
discontinuous linguistic expressions which 
are used to explicitly signal the discourse 
structure of a text. This paper describes 
efforts to improve an automatic tagging 
system which identifies and classifies 
discourse markers in Chinese texts by 
applying machine learning (ML) to the 
disambiguation of discourse markers, as an 
integral part of automatic text summarization 
via rhetorical structure. Encouraging results 
are reported. 
Keywords: discourse marker, Chinese 
corpus, rhetorical relation, automatic tagging, 
machine learning 
1 Introduction 
Discourse refers to any form of 
-based communication involving 
multiple sentences or utterances. The most 
important forms of discourse of interest to 
Natural Language Processing (NLP) are text 
and dialogue. The function of discourse 
analysis is to divide a text into discourse 
segments, and to recognize and re-construct 
the discourse structure of the text as intended 
by its author. 
Automatic text abstraction has received 
considerable attention (Paice 1990). Various 
systems have been developed (Chan et al. 
2000). Ono et al. (1994), T'sou et al. (1992) 
and Marcu (1997) focus on discourse 
structure in summarization using the 
Rhetorical Structure Theory (RST, Mann and 
Thompson 1986). The theory has been 
exploited in a number of computational 
systems (e.g. Hovy 1993). The main idea is 
to build a discourse tree where each node of 
the tree represents an RST relation. 
Summarization is achieved by trimming 
lmimportant sentences on the basis of the 
relative saliency or rhetorical relations. 
The SIFAS (Syntactic Marker based 
Full-Text Abstraction System) system has 
been implemented to use discourse markers 
in the automatic summarization of Chinese 
(T'sou et al. 1999). In this paper, we report 
our efforts to improve the SIFAS tagging 
system by applying machine learning 
techniques to disambiguation of discourse 
markers. C4.5 (Quirdan, 1993) is used in our 
system. 
2 Manual Tagging Process 
To tag the discourse markers, the 
following coding scheme is designed to 
encode Real Discourse Markers (RDM) 
appearing in the SIFAS corpus (T'sou et al. 
1998). We describe the z ~h discourse marker 
with a 7-tuple RDM; 
RDMi=< DM i, RRi, RPi, CTi, MNi, 
RNi, OT i >, where 
38 
DMi 
RP i : 
: 
MN~ : 
RNi : 
OT~ : 
the lexical item of the 
Discourse Marker, or the 
value'NULL'. 
the Rhetorical Relation in 
which DIVI~ is a constituent 
marker. 
the Relative Position of DMi. 
the Connection Type of RRi. 
the Discourse Marker 
Sequence Number. 
the Rhetorical Relation 
Sequence Number. 
the Order Type of RR~. The 
value of OTi can be 1, -1 or 0, 
denoting respectively the 
normal order, reverse order or 
irrelevance of the premise- 
consequence ordering of RR i . 
For apparent discourse markers that do 
not function as a real discourse marker in a 
text, a different coding scheme is used to 
encode them. We describe the i th apparent 
discourse marker using a 3-Tuple ADM~: 
ADMi =< LIi, *, SNi >, where 
LIi :the Lexical Item of the 
apparent discourse marker. 
SNi : the Sequence Number of the 
apparent discourse marker. 
In Chinese, discourse markers can be 
either words or phrases. To tag the SIFAS 
corpus, all discourse markers are organized 
into a discourse marker pair-rhetorical 
relation correspondence table. Part of the 
table is shown Table 1. 
To construct an automatic tagging 
system, let us first examine the sequential 
steps in the tagging process of a human 
tagger. 
S1. Written Chinese consists of rurming texts 
without word delimiters; the first step is 
is to segment the text into Chinese word 
sequences. 
$2. On the basis of a discourse marker list, 
we identify those words in the text 
which appear on the list as Candidate 
Discourse Markers (CDMs). 
$3. To winnow Real Discourse Markers 
(RDMs) and Apparent Discourse 
Markers (ADMs) from the CDMs, and 
encode the ADMs with a 3-tuple. 
$4. To encode the RDM with a 7-tuple 
according to a Discourse Marker Pair- 
Rhetorical Relation correspondence 
table. 
Relat- 
ion 
Adver- 
sativity 
Adver- 
sativity 
Causa- nty 
Causa- 
lity 
Front Back Con- 
nection 
Type 
Inter 
Intra 
Intra 1 
Intra -1 
Table 1 Discourse Marker Pair- 
Rhetorical Relation Table 
Order 
Type 
3 Automatic Tagging Process 
The identification of candidate discourse 
markers is based on a discourse marker list, 
which now contains 306 discourse markers 
plus a NULL marker. The markers are 
extracted from newspaper editorials of Hong 
Kong, Mainland China, Taiwan and 
Singapore. These markers constitute 480 
distinct discontinuous pairs that correspond 
to 25 rhetorical relations. In actual usage, 
some discourse marker pairs designate 
multiple rhetorical relations according to 
context. Some pairs can represent both 
INTER-sentence and INTRA-sentence 
relations. Thus the correspondence between 
the discourse marker pairs and the rhetorical 
relations is not single-valued. Some 
discourse marker pairs correspond to more 
than one rhetorical relation or connection 
type. We have 504 correspondences between 
the discourse marker pairs and the rhetorical 
relations. 
39 
In practice, one discontinuous 
constituent member of a marker pair is often 
omitted. We use the NULL marker to 
indicate the omission. In the 504 
correspondences, 244 of them are double 
constituent marker pairs, 260 are single 
constituent markers (i.e. One of the markers 
is NULL). And in the 244 double constituent 
markers, only 3 are not single-valued 
correspondences (one of" which is an 
INTER/INTRA relation, and can easily be 
distinguished.). Thus the tagging of the 244 
double constituent markers is basically a 
table searching process. But for the 260 
single constituent markers, the identity of the 
NULL marker is often difficult to determine. 
The SIFAS tagging system works in two 
modes: automatic and interactive (semi- 
automatic). The automatic tagging procedure 
is as follows: 
1. Data preparation: Input data files are 
modified according to the required 
format. 
2. Word segmentation: Because there are 
no delimiters between Chinese words in 
a text, words have to be extracted 
through a segmentation process. 
3. CDM identification 
4. Full-Marker RDM recognition 
5. ADM identification (first pass, 
deterministic) 
6. CDM feature extraction 
7. ADM identification (2nd pass, via ML) 
8. Tagging NuLL-marker CDM pairs (via 
ML) 
9.ADM and RDM sequencing, proof- 
reading, training data generation, and 
statistics 
The following principles are adopted by 
the tagging algorithm to resolve ambiguity in 
the process of matching discontinuous 
discourse markers: 
1. the principle of greediness: When 
matching a pair of discourse markers for 
a rhetorical relation, priority is given to 
the first matched relation from the left. 
2.the principle of locality: When 
matching a pair of discourse markers for 
a rhetorical relation, priority is given to 
the relation where the distance between 
its constituent markers is shortest. 
3.the principle of explicitness: When 
matching a pair of discourse markers for 
a rhetorical relation, priority is given to 
the relation where both markers are 
explicitly presented. 
4. the principle of superiority: When 
matching a pair of discourse markers for 
a rhetorical relation, priority is given to 
the inter-sentence relation whose back 
discourse marker matched with the first 
word of a sentence. 
5. the principle of Back-marker 
preference: This is applicable only to 
rhetorical relations where either the 
front or the back marker is absent, or to 
a NULL marker. In such cases, priority 
is given to the relation with the back 
marker present. 
Steps 1 to 6 and the five principles 
underlie the original naive tagger of the 
SIFAS system (T'sou et al. 1998), which also 
contains the system framework. 
4 Improvement 
4.1 Problems 
Many Chinese discourse markers have 
both discourse senses and alternate sentential 
senses in different context. For a human 
tagger, steps $3 and $4 in section 2 are not 
difficult because he/she can identify an 
ADM/RDM based on his/her text 
comprehension. However, for an automatic 
process, it is quite difficult to distinguish an 
ADM from an RDM if no syntactic/semantic 
information is available. 
Another problem is the location of 
NULL-Marker described above. Our earlier 
statistics showed some characteristics in the 
distance measured by punctuation marks. 
Statistics from 80 tagged editorials show that 
most of the relations are INTRA-Sentence 
relations (about 93%), about 70% of the 
INTRA RDM pairs have NULL markers. 
Most of these RDM pairs are separated by 
ONE comma (62%). These statistics show 
40 
the importance of the problems of 
positioning the NULL markers. 
The naive tagger partially solved the 
CDM discrimination and NULL marker 
location problems. Our experiment shows 
that about 45% of the ADMs can be 
correctly identified, and about 60% of the 
NULL markers can be correctly located one 
comma/period away from the current RDM. 
This leaves much room for improvement. 
One solution is to add a few rules 
according to previous statistics. The original 
naive tagger did not assume any knowledge 
of the statistics and behavioral patterns of 
discourse markers. From the error analysis, 
we extracted some additional rules to guide 
the classification and matching of the 
discourse markers. For example, one of the 
rules we extracted is: 
"A matching pair must be separated by 
at least two words or by punctuation 
marks". Using this rule, the following 
full marker matching error is avoided. 
< ~ ~ >< ~ ~ >< ~ x ~ >< 
./~ ,conjunction,Front, Intra,5,5,1>< ~ >< ~, 
conjunction, Back, Intra, 6,5,1><~>, <~t~><7~ 
±~x~><~><t$i~>, <~,*,7xff~>< 
X><~x<~><~x~><~x~_~ 
><~>0 
Another solution is to use 
• syntactic/semantic information through 
machine learning. 
4.2 C4.5 
Most empirical learning systems are 
given a set of pre-classified cases, each 
described by a vector of attribute values, and 
construct from them a mapping from 
attribute values to classes. C4.5 is one such 
system that learns decision-tree classifiers. It 
uses a divide-and-conquer approach to 
growing decision trees. The current version 
of C4.5 is C5.0 for Unix and See5 for 
Windows. 
Let attributes be denoted A={a~, a2, ..., 
a,,J, cases be denoted D={d 1, d2, ..., d J, and 
classes be denoted C={c, c 2, ..., cJ. For a 
set of cases D, a test 1q is a split of D based 
on attribute at. It splits D into mutually 
exclusive subsets D~, D 2, ..., D r These 
subsets of cases are single-class collections 
of cases. 
If a test T is chosen, the decision tree 
for D consists of a node identifying the test 
T, and one branch for each possible subset 
D~. For each subset D~, a new test is then 
chosen for further split. If D~ satisfies a 
stopping criterion, the tree for Dr is a leaf 
associated with the most frequent class in D~. 
One reason for stopping is that cases in D~ 
belong to one class. 
C4.5 uses arg max(gain(D,1)) or arg 
max(gain ratio(D,T)) to choose tests for 
split: 
k 
Info(D) = -~p(c,,D) * log2(p(c,,D)) 
i=I 
Split(D,T) = _LID, I. log2(~-~) 
i=l IDI 
Gain(D,T) = Info(D)- "J"'~.~'. Di I. Info(Di) 
i=l I DI 
Gain ratio(D, T) = gain(D, T) / Split(D, T) 
where, p(c~,D) denotes the proportion of 
cases in D that belong to the i th class. 
4.3 Application of C4.5 
Since using semantic information 
requires a comprehensive thesaurus, which is 
unavailable at present, we only use syntactic 
information through machine learning. 
The attributes used in the original 
SIFAS system include the candidate 
discourse marker itself, two words 
immediately to the left of the CDM, and two 
words immediately to the right of the CDM. 
The attribute names are F2, F1, CDM, B1, 
B2, respectively (T'sou et al, 1999). SIFAS 
only uses the Part Of Speech attribute of the 
neighboring words. This reflects to some 
degree the syntactic characteristics of the 
CDM. 
To reflect the distance characteristics, 
we add two other attributes: the number of 
discourse delimiters (commas, semicolons 
for INTRA-sentence relation, periods and 
41 
exclamation marks for INTER-sentence 
relation) before and after the current CDM, 
denoted Fcom and Boom, respectively. For 
the location of the NULL marker, we still 
add an actual number of delirniters Acorn. 
The order of these attributes is: CDM, 
F1, F2, B1, B2, Fcom, Boom Acorn for Null 
marker location, and CDM, F1, F2, B1, B2, 
Fcom, Bcom, IsRDM for CDM classification, 
where IsRDM is a Boolean value. 
The following are two examples of 
cases: 
9~: _N. ,?,q,a,a,7,1,1 for NULL marker 
location 
N~,d,?,u,?,l,0,F for CDM classificati 
on 
where "?" denotes that no corresponding 
word is at the position (beginning or end of 
sentence); a, d, q, and u are part-of-speech 
symbols in our segmentation dictionary, 
representing adjective, adverb, classifier, and 
auxiliary, respectively. 
The following are two examples of the 
rules generated by the C4.5. The first is a 
CDM classification rule, and the other is a 
NULL marker location rule. 
Rule 5: (11/1, lift 2.2) 
CDM = 
B1 =v 
Fcom > 0 
class T \[0.846\] 
which can be explained as: if the word after 
the CDM "~:" is a verb, and there is one 
comma in the sentence, before "~J:~:", then 
"~:" is an RDM. 
Rule 22: (1, lift 3.4) 
B2 = p 
Fcom > 1 
class 2 \[0.667\] 
which can be explained as: if the second 
word after the RDM is a preposition, and 
there is more then one commas before the 
current RDM, then the location of the NULL 
marker is two commas away from the RDM. 
4.4 Objects in the SIFAS system 
The objects in the new SIFAS tagging 
system are listed below. 
1. Dictionary Editor: for the update of 
word segmentation dictionary and the 
rhetorical relation table. 
2. Data Manager: for the modification of 
the input data (editorial texts) to 
conform with the required format. 
3. Word Segmenter: for the segmentation 
of the original texts, and the recognition 
of CDMs. 
4. RDM Tagger: The initial identification 
of RDMs is a table searching process. 
All those full-marker pairs are identified 
as rhetorical relations according to the 
principles described above. For those 
Null-marker pairs, the location of the 
Null maker is left to the rule interpreter. 
5. ADM Tagger: The identification of 
ADMs is also a table searching process, 
because, without other 
syntactic/semantic information, the only 
way to identify ADMs from the CDMs 
is to find out that the CDM cannot form 
a valid pair with any other CDMs 
(including the NULL marker) to 
correspond to a rhetorical relation. 
6. CDM Feature Extractor: For those 
untagged CDMs, the classification is 
carried out through C4.5. The Feature 
Extractor extracts syntactic information 
about the current CDM and send it to the 
Rule Interpreter (see below). 
7. Rule Interpreter: C4.5 takes feature data 
file as the input to construct a classifier, 
and the rules formed are stored in an 
output file. The rule interpreter reads 
this output file and applies the rules to 
classify the CDMs. In our system, The 
Rule Interpreter functions as a NULL 
Marker Locator and a CDM classifier. 
8. Sequencer: for the rearrangement of 
RDM and ADM order number. In the 
rearranging process, the Sequencer also 
extracts statistical information for 
analysis. 
9. Interaction Recorder: for the recording 
of user interaction information for 
42 
statistics use. 
10. Data Retriever: for data retrieval and 
browsing. 
5 Evaluation 
In order to evaluate the effectiveness of 
the tagging system in terms of the percentage 
of discourse markers that can be tagged 
correctly, we have chosen 80 tagged 
editorials from Ming Pao, a Chinese 
newspaper of Hong Kong, in the duration 
from December 1995 to January 1996 to 
form a training data set. Then we randomly 
selected 20 editorials from Mainland China 
and Hong Kong newspapers for the system 
to tag automatically, and then manually 
checked the results. 
The total CDMs in the training data set 
is 4764, in which 2116 are RDMs and 2648 
are ADMs. The distribution of INTER- 
sentence relations, INTRA-sentence relations, 
and NULL marker pairs is shown below. 
Total 
Relations 
Inter- 
Sentence 
Relations 
Intra- 
Sentence 
Relations 
Relations 
with 
NULL 
marker 
pair 
1589 98 1491 1062 
100% 6.17% 93.83% 66.83% 
Table 2 Distribution of INTER-/INTRA- 
sentence relations, 
and NULL marker pairs 
Our evaluation is based on counting the 
number of discourse markers that are 
correctly and incorrectly tagged. 
The total CDMs in the test data set is 
1134, in which 563 are RDMs and 571 are 
ADMs. The distribution of INTER-sentence 
relations, INTRA-sentence relations, and 
NULL marker pairs in the test data set is 
shown in Table 3. 
Total 
Relations 
Inter- 
Sentence 
Relations 
Intra- 
Sentence 
Relations 
Relations 
with 
NULL 
marker 
pair 
424 23 401 285 
100% 5.42% 94.58% 67.22% 
Table 3 Distribution of INTER-/INTRA- 
sentence relations, and NULL marker 
pairs in testing data set 
451 399 11 1 65 3 
Table 4 Test Results 
From the test results shown in Table 4, 
we can see that most of the errors are caused 
by the misclassification of the CDMs. An 
example of Other errors is shown below. 
The following sentence is from an editorial 
of People's Daily. 
< ~ ~17 >< ~ ~.~ >< ~ ~ > , 
<NULL,sufficienc y, Front, Intra, O,81,1x -- ~ ~ff \[\] 
><~ ><~. ~lJ><~.~.>< \[\] ~>, <~><~iA><-- 
+~ \[\]><~><)~lJ>, <~><~iA><~>< 
~,*,80><~ \[\]><1~><--~52">, <~~>< 
~,sufticiency, Back,Intra,81,81,1 < ~ ~ x :~ x 
~.><~><~>, <~ ~><~ :~ ><~,*,SZ><~ l~J, 
><~>° 
In the above sentence, the first "R" is 
matched with the NULL marker, but the 
second "R" is left as an ADM. This causes 
an "Other error" and an "ADM/RDM 
classification error". 
The Gross Accuracy (GA) as defined in 
T'sou et al. (1999) is: 
GA = correctly tagged discourse 
markers / total number of discourse markers 
= 95.38% 
This greatly improves the performance 
compared with the original GA = 68.89%. 
The overgeneration problem (tagged 415, 
actual 424) is caused by the mismatch of 
CDMs as RDM pairs, or by the 
43 
misclassification of CDMs as RDMs. 
Following are two examples. 
< ~\[I ~ ,sufficieney, Front, Intra,54,54,1x ~ ~1"\] >_< 
~\[1- ~><~,*,56x~xff ~ >, <~A.x~x~ :t: 
>< ~ > , < ~1~. ,sufficieney, Back,Intra,57,54,1>_< 
,*,58><~ ~x~ ~><~><:~ ~x~ ~x 
><~I l~.l><~x(~ ~ x~,*,59><~ A.x~ ~ 
><:t:~><--~>° 
In this example, "~tl ~" could have 
matched <:~,*.55>, <,~,*,56>, or <~,*,58>. 
Only the <:~,*,55> and the <~,*,58> can be 
eliminated from the candidates according to 
the "simple rules" mentioned in section 4.1. 
The system has to choose from <~,*,56> and 
<}J~,*,57> to match with "~zn~'. Luckily, 
the system has given a right choice here. 
< -- ~" ~ \[\] >< ~ ,conjunction,Front, Intra,46, 
46,1><~~><~><~r~> , <NULL, 
conjunction,Front, Intra,0,49,1 ><-- f" ~ \[\] ><)~ ~ 
>< ~,~ ,conjunction,Back, Intra,47,46,1>< 
~i~,*,48>< ~ ~><1~ \]\]~><~E ~><~><~ ~>< 
~ \[\] A.><~><~ ~ ><th~> , < 
,eonjunction,Baek,Intra,49,49,1>< ~ ~tJ >< ~ 
><±x~ ~><\[\] ~><~n><~,*,50><l~# 
\[\]><~ I~x~x\[\] g,~><~ ~>< ~ ><Lib>. 
The two "~" are misclassified as RDMs, 
and causes a mismatch of RDM pair. Such 
errors are difficult to avoid for an automatic 
system. Without further syntactic/semantic 
analysis, we can only hope for the ML 
algorithm to give us a solution from more 
training data. 
6 Conclusion 
In order to study discourse markers for 
use in the automatic summarization of 
Chinese text, we have designed and 
implemented the SIFAS system. In this 
paper, we have focused on the problems of 
NULL marker location and the classification 
of RDMs and ADMs. A study on applying 
machine learning techniques to discourse 
marker disambiguation is conducted. C4.5 is 
used to generate decision tree classifiers. Our 
results indicate that machine learning is an 
effective approach to improving the accuracy 
of discourse marker tagging. For interactive 
use of the system, if we set a threshold for 
the rule precision and only display those low 
precision rules for interactive selection, we 
can greatly speed up the semi-automatic 
tagging process. 

References 
Chart S., Lai T., Gao W. J. and T'sou B. K. 
(2000) "Mining Discourse Markers for 
Chinese Textual Summarization." In 
Proceedings of the Sixth Applied Natural 
Language Processing Conference and the 
North American Chapter of the 
Association for Computational Linguistics. 
Workshop on Automatic Summarization, 
Seattle, Washington, 29 April to 3 May, 
2000. 
Grosz B.J. and Sidner C. (1986) "Attention, 
Intention, and the Structure of Discourse," 
Computational Linguistics 12(3): 175-204. 
Hirst G. (1981) "Discourse Oriented 
Anaphoral Resolution in Natural Language 
Understanding: A Review." Computational 
Linguistics 7(2): 85-98. 
Hovy E. (1993) "Automated Discourse 
Generation using Discourse Structure 
Relations." Artificial Intelligence 63: 341- 
385. 
Hwang C. H. and Schubert L. K. (1992) 
"Tense Trees as the 'Fine Structure' of 
Discourse." In Proc. 30th Annual Meeting, 
Assoc. for Computational Linguistics, pp. 
232-240. 
Lin H. L., T'sou B. K., H. C. Ho, Lai T., Lun 
C., C. K. Choi and C.Y. Kit. (1991) 
"Automatic Chinese Text Generation 
Based on Inference Trees." In Proe. of 
ROCLING Computational Linguistic 
Conference IV, Taipei, pp. 215-236. 
Litman D. J. and Allen J. (1990) "Discourse 
Processing and Commonsense Plans." In 
Cohen et al.(ed.) Intentions in 
Communications, pp. 365-388. 
Mann W. C. and Thompson S. A (1988) 
"Rhetorical Structure Theory: Towards a 
Functional Theory of Text Organization." 
44 
• Text 8(3): 243-281. 
Marcu D. (1997) "From Discourse Structures 
to Text Summaries." In Proceedings of the 
ACL/EACL'97 Workshop on Intelligent 
Scalable Text Summarization, Spain, pp. 
82-88. 
McKeown K. and Radev D. (1995) 
"Summaries of Multiple News Articles." 
In Proceedings of the 18th Annual 
International ACM SIGIR Conference on 
Research and Development in Information 
Retrieval, Seattle, pp. 74-82. 
Ono K., Surnita K. and S. Miike. (1994) 
"Abstract Generation based on Rhetorical 
Structure Extraction." In Proceedings of 
International Conference on 
Computational Linguistics, Japan, pp. 344- 
348. 
Paice C. D. (1990) "Constructing Literature 
Abstracts by Computer: Techniques and 
Prospects." Information Processing and 
Management 26(1): 171-186. 
Qulnlan J. Ross (1993) "C4.5 Programs for 
Machine Learning." San Mateo, CA: 
Morgan Kaufmann. 
T'sou B. K., Ho H. C., Lai B. ¥., Lun C. and 
Lin H. L. (1992) "A Knowledge-based 
Machine-aided System for Chinese Text 
Abstraction." In Proceedings of 
International Conference on 
Computational Linguistics, France, pp. 
1039-1042. 
T'sou B. K., Gao W. J., Lin H. L., Lai T. B. 
Y. and Ho H. C. (1999) "Tagging 
Discourse Markers: Towards a Corpus 
based Study of Discourse Marker Usage in 
Chinese Text" In Proceedings of the 18th 
International Conference on Computer 
Processing of Oriental Languages, March 
1999, Japan, pp. 391-396. 
T'sou B. K., Lin H. L., Ho H. C., Lai T. and 
Chan T. (1996) "Automated Chinese Full- 
text Abstraction Based on Rhetorical 
Structure Analysis." Computer Processing 
of Oriental Languages 10(2): 225-238. 
Tsou, B.K., et al., 1998: ~1~, ~, 
i~~1-~3-~~ ~ ~", 
ICCIP'98, Beijing, Nov. 18-20, 1998. 
