Applying Coreference to Improve Name Recognition 
Heng JI and Ralph GRISHMAN 
Department of Computer Science 
New York University 
715 Broadway, 7
th
 Floor 
New York, NY 10003, U.S.A. 
   hengji@cs.nyu.edu,  grishman@cs.nyu.edu 
 
Abstract 
We present a novel method of applying the 
results of coreference resolution to improve 
Name Recognition for Chinese.  We consider 
first some methods for gauging the confidence 
of individual tags assigned by a statistical 
name tagger.  For names with low confidence, 
we show how these names can be filtered 
using coreference features to improve 
accuracy.  In addition, we present rules which 
use coreference information to correct some 
name tagging errors.  Finally, we show how 
these gains can be magnified by clustering 
documents and using cross-document 
coreference in these clusters.  These combined 
methods yield an absolute improvement of 
about 3.1% in tagger F score. 
 
1 Introduction 
The problem of name recognition and 
classification has been intensively studied since 
1995, when it was introduced as part of the MUC-
6 Evaluation (Grishman and Sundheim, 1996).  A 
wide variety of machine learning methods have 
been applied to this problem, including Hidden 
Markov Models (Bikel et al. 1997), Maximum 
Entropy methods (Borthwick et al. 1998, Chieu 
and Ng 2002), Decision Trees (Sekine et al. 1998), 
Conditional Random Fields (McCallum and Li 
2003), Class-based Language Model (Sun et al. 
2002), Agent-based Approach (Ye et al. 2002) and 
Support Vector Machines. However, the 
performance of even the best of these models
1
 has 
been limited by the amount of labeled training data 
available to them and the range of features which 
they employ.  In particular, most of these methods 
classify an instance of a name based on the 
information about that instance alone, and very 
local context of that instance – typically, one or 
                                                   
1
  The best results reported for Chinese named entity 
recognition, on the MET-2 test corpus, are 0.92 to 0.95   
F-measure for the different name types (Ye et al. 2002). 
two words preceding and following the name.  If a 
name has not been seen before, and appears in a 
relatively uninformative context, it becomes very 
hard to classify. 
We propose to use more global information to 
improve the performance of name recognition.  
Some name taggers have incorporated a name 
cache or similar mechanism which makes use of 
names previously recognized in the document.  In 
our approach, we perform coreference analysis and 
then use detailed evidence from other phrases in 
the document which are co-referential with this 
name in order to disambiguate the name.  This 
allows us to perform a richer set of corrections 
than with a name cache.  We then go one step 
further and process similar documents containing 
instances of the same name, and combine the 
evidence from these additional instances.  At each 
step we are able to demonstrate a small but 
consistent improvement in named entity 
recognition. 
The rest of the paper is organized as follows. 
Section 2 briefly describes the baseline name 
tagger and coreference resolver used in this paper. 
Section 3 considers methods for assessing the 
confidence of name tagging decisions.  Section 4 
examines the distribution of name errors, as a 
motivation for using coreference information. 
Section 5 shows the coreference features we use 
and how they are incorporated into a statistical 
name filter.  Section 6 describes additional rules 
using coreference to improve name recognition. 
Section 7 provides the flow graph of the improved 
system.  Section 8 reports and discusses the 
experimental results while Section 9 summarizes 
the conclusions. 
2 Baseline Systems 
The task we consider in this paper is to identify 
three classes of names in Chinese text:  persons 
(PER), organizations (ORG), and geo-political 
entities (GPE).  Geo-political entities are locations 
which have an associated government, such as 
cities, states, and countries.
2
  Name recognition in 
Chinese poses extra challenges because neither 
capitalization nor word segmentation clues are 
explicitly provided, although most of the 
techniques we describe are more generally 
applicable. 
Our study builds on an extraction system 
developed for the ACE evaluation, a multi-site 
evaluation of information extraction organized by 
the U.S. Government.  Following ACE 
terminology, we will use the term mention to refer 
to a name or noun phrase of one of the types of 
interest, and the term entity for a set of coreferring 
mentions.  We briefly describe in this section the 
baseline Chinese named entity tagger, as well as 
the coreference system, used in our experiments. 
2.1 Chinese Name Tagger 
Our baseline name tagger consists of an HMM 
tagger augmented with a set of post-processing 
rules.  The HMM tagger generally follows the 
NYMBLE model (Bikel et al, 1997), but with a 
larger number of states (12) to handle name 
prefixes and suffixes, and transliterated foreign 
names separately.  It operates on the output of a 
word segmenter from Tsinghua University.  It uses 
a trigram model with dynamic backoff.  The post-
processing rules correct some omissions and 
systematic errors using name lists (for example, a 
list of all Chinese last names; lists of organization 
and location suffixes) and particular contextual 
patterns (for example, verbs occurring with 
people’s names).  They also deal with 
abbreviations and nested organization names. 
 
2.2 Chinese Coreference Resolver 
For this study we have used a rule-based 
coreference resolver.  Table 1 lists the main rules 
and patterns used.  We have extensive rules for 
name-name coreference, including rules specific to 
the particular name types.  For these experiments, 
we do not attempt to resolve pronouns, and we 
only resolve names with nominals when the name 
and nominal appear in close proximity in a specific 
structure, as listed in Table 1. 
We have used the MUC coreference scoring 
metric (Vilain et al, 1995) to evaluate this resolver, 
excluding all pronouns and limiting ourselves to 
noun phrases of semantic type PER, ORG, and 
GPE.  Using a perfect (hand-generated) set of 
mentions, we obtain a recall of 82.7% and 
precision of 95.1%, for an F score of 88.47%.  
                                                   
2
 This class is used in the U.S. Government’s ACE 
evaluations;  it excludes locations without governments, 
such as bodies of water and mountains. 
Using the mentions generated by our extraction 
system, we obtain a recall of 74.3%, a precision of 
84.5%, and an F score of 79.07%.
3
  
3 Confidence Measures 
In order to decide when we need to rely on 
global (coreference) information for name tagging, 
we want to have some assessment of the 
confidence that the name tagger has in individual 
tagging decisions.  In this paper, we use two tools 
to reach this goal.  The first method is to use three 
manually built proper name lists which include 
common names of each type (selected from the 
high frequency names in the user query blog of 
COMPASS, a Chinese search engine, and name 
lists provided by Linguistic Data Consortium; the 
PER list includes 147 names, the GPE list 226 
names, and the ORG list 130 names).  Names on 
these lists are accepted without further review. 
The second method is to have the HMM tagger 
compute a probability margin for the identification 
of a particular name as being of a particular type.  
Scheffer et al. (2001) used a similar method to 
identify good candidates for tagging in an active 
learner.  During decoding, the HMM tagger seeks 
the path of maximal probability through the Viterbi 
lattice.  Suppose we wish to evaluate the 
confidence with which words w
i
, …, w
j
 are 
identified as a name of type T.  We compute 
 
Margin (w
i
,…, w
j
; T) =  log P
1
 – log P
2
 
 
Here P
1
 is the maximum path probability and P
2
 is 
the maximum probability among all paths for 
which some word in w
i
, …, w
j
 is assigned a tag 
other than T. 
A large margin indicates greater confidence in 
the tag assignment.  If we exclude names tagged 
with a margin below a threshold, we can increase 
the precision of name tagging at some cost in recall.  
Figure 1 shows the trade-off between margin 
threshold and name recognition performance. 
Names with a margin over 3.0 are accepted on this 
basis. 
                                                   
3
 In our scoring, we use the ACE keys and only score 
mentions which appear in both the key and system 
response.  This therefore includes only mentions 
identified as being in the ACE semantic categories by 
both the key and the system response.  Thus these 
scores cannot be directly compared against coreference 
scores involving all noun phrases. 
85
87
89
91
93
95
97
99
012345678910112
Threshold
P
r
e
c
i
s
i
on(
%
)
 
Figure 1: Tradeoff between Margin Threshold and 
name recognition performance 
 
4   Distribution of Name Errors 
We consider now names which did not pass the 
confidence measure tests: names not on the 
common name list, which were tagged with a 
margin below the threshold.  We counted the 
accuracy of these “obscure” names as a function of 
the number of mentions in an entity; the results are 
shown in Table 2. 
The table shows that the accuracy of name 
recognition increases as the entity includes more 
mentions.  In other words, if a name has more 
coref-ed mentions, it is more likely to be correct. 
This also provides us a linguistic intuition: if 
people mention an obscure name in a text, they 
tend to emphasize it later by repeating the same 
name or describe it with nominal mentions. 
The table also indicates that the accuracy of 
single name entities (singletons) is much lower 
than the overall accuracy.  So, although they 
constitute only about 10% of all names, increasing 
their accuracy can significantly improve overall 
performance.  Coreference information can play a 
great role here.  Take the 157 PER singletons as an 
example; 56% are incorrect names. Among these 
incorrect names, 73% actually belong to the other 
two name types.  Many of these can be easily fixed 
by searching for coreference to other mentions 
without type restriction.  Among the correct names, 
71% can be confirmed by the presence of a title 
word or a Chinese last name.  From these 
observations we can conclude that without strong 
confirmation features, singletons are much less 
likely to be correct names. 
5 Incorporating Coreference Information 
into Name Recognition 
We make use of several features of the 
coreference relations a name is involved in; the 
features are listed in Table 3.  Using these features, 
we built an independent classifier to predict if a 
name identified by the baseline name tagger is 
correct or not. (Note that this classifier is trained 
on all name mentions, but during test only 
‘obscure’ names which failed the tests in section 3 
are processed by this classifier.) Each name 
corresponds to a feature vector which consists of 
the factors described in Table 3.  The PER context 
words are generated from the context patterns 
described in  (Ji and Luo, 2001).  We used a 
Support Vector Machine to implement the 
classifier, because of its state-of-the-art 
performance and good generalization ability.  We 
used a polynomial kernel of degree 3. 
6 Name Rules based on Coreference 
Besides the factors in the above statistical model, 
additional coreference information can be used to 
filter and in some cases correct the tagging 
produced by the HMM.  We developed the 
following rules to correct names generated by the 
baseline tagger. 
6.1 Name Structure Errors 
Sometimes the Name tagger outputs names 
which are too short (incomplete) or too long.  We 
can make use of the relation among mentions in 
the same entity to fix them.  For example, nested 
ORGs are traditionally difficult to recognize 
correctly.  Errors in ORG names can take the 
following forms: 
 
(1) Head Missed. Examples: “中国艺术（团） / 
Chinese Art (Group)”, “中国学生（会） / Chinese 
Student (Union)”, “俄罗斯核动力（所） / Russian 
Nuclear Power (Instituition)” 
 
Rule 1: If an ORG name x is coref-ed with other 
mentions with head y (an ORG suffix), and in the 
original text x is immediately followed by y, then 
tag xy instead of x; otherwise discard x. 
 
(2) Modifier Missed. Rule 1 can also be used to 
restore missed modifiers. For example, “（爱丁
堡）大学  / (Edinburgh) University”; “（鹏程）有
限公司  / (Peng Cheng) Limited Corporation”, and 
some incomplete translated PER names such as 
“（巴）勒斯坦  / (Pa)lestine”. 
 
(3) Name Too Long 
Rule 2: If a name x has no coref-ed mentions 
but part of it, x', is identical to a name in another 
entity y, and y includes at least two mentions; then 
tag x' instead of x. 
.
 
Rule Type Rule Description 
Ident(i, j) Mention
i
 and Mention
j
 are identical 
Abbrev(i, j) Mention
i
 is an abbreviation of Mention
j
 
Modifier(i, j) Mention
j
 = Modifier + “de” + Mention
i
 
 
 
All 
Formal(i, j) Formal and informal ways of referring to the same entity 
(Ex. “美国国防部  / American Defense Dept. &  
五角大楼 / Pentagon”) 
Substring(i, j) Mention
i
 is a substring of Mention
j
   
PER Title(i, j) Mention
j
 = Mention
i
 + title word; or 
Mention
j
 = LastName + title word 
ORG Head(i, j) Mention
i
 and Mention
j
 have the same head 
Head(i, j) Mention
i
 and Mention
j
 have the same head 
Capital(i, j) Mention
i
: country name; 
Mention
j
: name of the capital of this country 
Applied in restricted context. 
 
 
 
 
 
 
 
 
Name & 
Name 
 
 
 
GPE 
Country(i, j) Mention
i
 and Mention
j
 are different names referring to the same 
country.  
(Ex. “中国  / China & 华夏  / Huaxia & 共和国  / Republic”) 
RSub(i, j) Name
i
 is a right substring of Nominal
j
 
Apposition(i, j) Nominal
j
 is the apposite of Name
i
 
 
All 
Modifier2(i, j) Nominal
j
 = Determiner/Modifier + Name
i
/ head 
 
 
Name & 
Nominal 
 
GPE 
Ref(i, j) Nominal
j
 = Name
i
 + GPE Ref Word  
(examples of GPE Ref Word: “方面  / Side”, “政府 /Government”, 
“共和国  / Republic”, “自治政府 / Municipality”) 
IdentN(i, j) Nominal
i
 and Nominal
j
 are identical Nominal& 
Nominal 
All 
Modifier3(i, j) Nominal
j
 = Determiner/Modifier + Nominal
i
 
 
Table1: Main rules used in the Coreference Resolver 
 
Number of mentions 
per entity 
Name Type 
1 
 
2 3 4 5 6 7 8 >8 
PER 43.94 87.07 91.23 87.95 91.57 91.92 94.74 92.31 97.36 
GPE 55.81 88.8 96.07 100 100 100 100 95.83 97.46 
ORG 64.71 80.59 89.47 94.29 100 100 -- -- 100 
 
Table 2 Accuracy(%) of ‘obscure’ name recognition 
 
 Factor Description 
Coreference Type 
Weight 
Average of weights of coreference relations for which this mention 
is antecedent:  0.8 for name-name coreference; 0.5 for apposition;  
0.3 for other name-nominal coreference 
First 
Mention 
Is first name mention in the entity 
Head Includes head word of name 
Idiom Name is part of an idiom 
PER context For PER Name, has context word in text 
PER title For PER Name, includes title word 
 
 
Mention 
Weight 
ORG suffix For ORG Name, includes suffix word 
Entity Weight Number of mentions in entity / total number of mentions in all 
entities in document which include a name mention  
 
Table 3 Coreference factors for name recognition 
6.2 Name Type Errors 
Some names are mistakenly recognized as other 
name types.  For example, the name tagger has 
difficulty in distinguishing transliterated PER 
name and transliterated GPE names. 
To solve this problem we designed the 
following rules based on the relation among 
entities. 
Rule 3: If name
i
 is recognized as type1, the 
entity it belongs to has only one mention; and 
name
j
 is recognized as type2, the entity it belongs 
to has at least two mentions; and name
i
 is identical 
with name
j
 or name
i
 is a substring of name
j
, then 
correct type1 to type2. 
 
For example, if “ 克里姆林 / Kremlin” is 
mistakenly identified as PER, while “克里姆林宫  
/ Kremlin Palace” is correctly identified as ORG, 
and in coreference results, “克里姆林  / Kremlin” 
belongs to a singleton entity, while “克里姆林宫  / 
Kremlin Palace” has coref-ed mentions, then we 
correct the type of “克里姆林  / Kremlin” to ORG.  
 
Another common mistake gives rise to the 
sequence “PER+title+PER”, because our name 
tagger uses the title word as an important context 
feature for a person name (either preceding or 
following the title).  But this is an impossible 
structure in Chinese.  We can also use coreference 
information to fix it. 
Rule 4: If “PER+title+PER” appears in the 
name tagger’s output,  then we discard the PER 
name with lower coref certainty; and check 
whether it is coref-ed to other mentions in a GPE 
entity or ORG entity; if it is, correct the type. 
Using this rule we can correctly identify “[斯里
兰卡  / Sri Lanka GPE] 总理  / Premier [班达拉耐
克  / Bandaranaike PER]”, instead of “[斯里兰卡  / 
Sri Lanka PER] 总理  / Premier [班达拉耐克  / 
Bandaranaike PER]”. 
 
6.3 Name Abbreviation Errors 
Name abbreviations are difficult to recognize 
correctly due to a lack of training data.  Usually 
people adopt a separate list of abbreviations or 
design separate rules (Sun et al. 2002) to identify 
them.  But many wrong abbreviation names might 
be produced.  We find that coreference 
information helps to select abbreviations. 
Rule 5: If an abbreviation name has no coref-ed 
mentions and it is not adjacent to another 
abbreviation (ex. “中 /China 美 /America”), then 
we discard it. 
 
7 System Flow 
Combining all the methods presented above, the 
flow of our final system is shown in Figure 2: 
 
 
Figure 2  System Flow 
 
8 Experiments 
8.1 Training and Test Data 
For our experiments, we used the Beijing 
University Insititute of Computational Linguistics 
corpus – 2978 documents from the People’s Daily 
in 1998, one million words with name tags – and 
the training corpus for  the 2003 ACE evaluation, 
223 documents.  153 of our ACE documents were 
used as our test set.
4
  The 153 documents 
contained 1614 names.  Of the system-tagged 
names, 959 were considered ‘obscure’:  were not 
on a name list and had a margin below the 
threshold.  These were the names to which the 
rules and classifier were applied.  We ran all the 
following experiments using the MUC scorer.  
                                                   
4
 The test set was divided into two parts, of 95 
documents and 58 documents.  We trained two name 
tagger and classifier models, each time using one part 
of the test set along with all the other documents, and 
evaluated on the other part of the test set.  The results 
reported here are the combined results for the entire 
test set. 
Input 
 
Name 
tagger 
Nominal 
mention 
tagger 
Coreference 
Resolver 
Coreference 
Rules to fix  name 
errors 
SVM classifier to select 
correct names using 
coreference features 
Output 
8.2 Overall Performance Comparison 
Table 4 shows the performance of the baseline 
system; Table 5 the system with rule-based 
corrections; and Table 6 the system with both 
rules and the SVM classifier. 
 
Name Precision Recall F 
PER 90.9 88.2 89.5 
GPE 82.3 90.8 86.3 
ORG 92.1 91.8 91.9 
ALL 87.8 90.5 89.1 
 
Table 4 Baseline Name Tagger 
 
Name Precision Recall F 
PER 93.3 87.5 90.3 
GPE 83.5 90.4 86.8 
ORG 90.9 92.1 91.5 
ALL 88.5 90.3 89.4 
 
Table 5 Results with Coref Rules Alone 
 
Name Precision Recall F 
PER 95.7 84.4 89.7 
GPE 88.0 91.7 89.8 
ORG 94.5 91.2 92.8 
ALL 92.2 89.6 90.9 
 
Table 6 Results for Single Document System 
 
The gains we observed from coreference within 
single documents suggested that further 
improvement might be possible by gathering 
evidence from several related documents.
5
  We 
did this in two stages.  First, we clustered the 153 
documents in the test set into 38 topical clusters.  
Most (29) of the clusters had only two documents;  
the largest had 28 documents.  We then applied 
the same procedures, treating the entire cluster as 
a single document.  This yielded another 1.0% 
improvement in overall F score (Table 7). 
The improvement in F score was consistent for 
the larger clusters (3 or more documents):  the F 
score improved for 8 of those clusters and 
remained the same for the 9
th
.  To heighten the 
multi-document benefit, we took 11 of the small 
                                                   
5
 Borthwick (1999) did use some cross-document 
information across the entire test corpus, maintaining 
in effect a name cache for the corpus, in addition to one 
for the document.  No attempt was made to select or 
cluster documents. 
(2 document clusters) and enlarged them by 
retrieving related documents from 
sina.com.cn.  In total, we added 52 texts to 
these 11 clusters.  The net result was a further 
improvement of 0.3% in F score (Table 8).
6
 
 
 
Name Precision Recall F 
PER 93.3 86.8 90.5 
GPE 95.2 90.0 92.5 
ORG 92.9 91.7 92.3 
ALL 93.8 90.1 91.9 
 
Table 7 Results for Mutiple Document System 
 
Name Precision Recall F 
PER 94.7 87.1 90.7 
GPE 95.6 89.6 92.5 
ORG 95.8 90.3 93.0 
ALL 95.4 89.2 92.2 
 
Table 8 Results for Mutiple Document System 
               with additional retrieved texts 
 
8.3 Contribution of Coreference Features 
Since feature selection is crucial to SVMs, we 
did experiments to determine how precision 
increased as each feature was added.  The results 
are shown in Figure 3.  We can see that each 
feature in the SVM helps to select correct names 
from the output of the baseline name tagger, 
although some (like FirstMention) are more 
crucial than others. 
 
87
88
89
90
91
92
93
B
as
e
l
in
e
Ed
g
eT
y
p
e
F
i
rs
t
M
e
n
ti
o
n
He
ad
Id
io
m
P
e
rC
o
n
t
ex
t
P
er
T
i
tl
e
O
rg
S
u
ff
ix
E
n
tit
y
W
e
i
g
h
t
Feature
P
r
ecis
i
o
n
(
%
)
 
Figure 3  Contributions of features 
 
                                                   
6
 Scores are still computed on the 153 test 
documents ;  the retrieved documents are excluded 
from the scoring. 
8.4 Comparison to Cache Model 
Some named entity systems use a name cache, 
in which tokens or complete names which have 
been previously assigned a tag are available as 
features in tagging the remainder of a document.  
Other systems have made a second tagging pass 
which uses information on token sequences 
tagged in the first pass (Borthwick 1999), or have 
used as features information about features 
assigned to other instances of the same token 
(Chieu and Ng 2002).  Our system, while more 
complex, makes use of a richer set of global 
features, involving the detailed structure of 
individual mentions, and in particular makes use 
of both name – name and name – nominal 
relations. 
 
We have compared the performance of our 
method (applied to single documents) with a 
voted cache model, which takes into account the 
number of times a particular name has been 
previously assigned each type of tag: 
 
System Precision Recall F 
baseline 88.8 90.5 89.1 
 voted cache 87.6 92.8 90.1 
current 92.2 89.6 90.9 
 
Table 9.  Comparison with voted cache 
 
Compared to a simple voted cache model, our 
model provides a greater improvement in name 
recognition F score; in particular, it can 
substantially increase the precision of name 
recognition.  The voted cache model can recover 
some missed names, but at some loss in precision. 
9 Conclusions and Future Work 
In this paper, we presented a novel idea of 
applying coreference information to improve 
name recognition.  We used both a statistical filter 
based on a set of coreference features and rules 
for correcting specific errors in name recognition.  
Overall, we obtained an absolute improvement of 
3.1% in F score.  Put another way, we were able 
to eliminate about 60% of erroneous name tags 
with only a small loss in recall. 
The methods were tested on a Chinese name 
tagger, but most of the techniques should be 
applicable to other languages.  More generally, it 
offers an example of using global and cross-
document information to improve local decisions 
for information extraction.  Such methods will be 
important for breaking the ‘performance ceiling’ 
in many areas of information extraction. 
In the future, we plan to experiment with 
improvements in coreference resolution (in 
particular, adding pronoun resolution) to see if we 
can obtain further gains in name recognition.  We 
also intend to explore the production of multiple 
tagging hypotheses by our statistical name tagger, 
with the alternative hypotheses then reranked 
using global information.  This may allow us to 
replace some of our hand-coded error-correction 
rules with corpus-trained methods. 
10 Acknowledgements 
This research was supported by the Defense 
Advanced Research Projects Agency as part of the 
Translingual Information Detection, Extraction 
and Summarization (TIDES) program, under 
Grant N66001-001-1-8917 from the Space and 
Naval Warfare Systems Center San Diego, and by 
the National Science Foundation under Grants 
IIS-0081962 and 0325657.  This paper does not 
necessarily reflect the position or the policy of the 
U.S. Government. 

References  
Daniel M. Bikel, Scott Miller, Richard Schuartz, 
and Ralph Weischedel. 1999. Nymble: a high-
performance Learning Name-finder.  Proc. Fifth 
Conf. On Applied Natural Language Processing, 
Washington, D.C. 
Andrew Borthwick.  1999. A Maximum Entropy 
Approach to Named Entity Recognition.  Ph.D. 
Dissertation, Dept. of Computer Science, New 
York University. 
Andrew Borthwick, John Sterling, Eugene 
Agichtein, and Ralph Grishman. 1998. 
Exploiting Diverse Knowledge Sources via 
Maximum Entropy in Named Entity 
Recognition. Proc. Sixth Workshop on Very 
Large Corpora, Montreal, Canada. 
Hai Leong Chieu and Hwee Tou Ng. 2002.  
Named Entity Recognition: A Maximum 
Entropy Approach Using Global Information.  
Proc.: 17th Int’l Conf. on Computational 
Linguistics (COLING 2002), Taipei, Taiwan. 
Ralph Grishman and Beth Sundheim.  1996. 
Message understanding conference - 6: A brief 
history. Proc. 16th Int’l Conference on 
Computational Linguistics (COLING 96), 
Copenhagen. 
Heng Ji, Zhensheng Luo, 2001. A Chinese Name 
Identifying System Based on Inverse Name 
Frequency Model and Rules. Natural Language 
Processing and Knowledge Engineering 
(NLPKE) Mini Symposium of 2001 IEEE 
International Conference on Systems, Man, and 
Cybernetics (SMC2001) 
Andrew McCallum and Wei Li.  2003.  Early 
results for Named Entity Recognition With 
Conditional Random Fields, Feature Induction, 
and Web-Enhanced Lexicons. Proc. Seventh 
Conf. on Computational Natural Language 
Learning (CONLL-2003), Edmonton, Canada. 
Tobias Scheffer, Christian Decomain, and Stefan 
Wrobel. 2001. Active Hidden Markov Models 
for Information Extraction. Proc. Int’l 
Symposium on Intelligent Data Analysis (IDA-
2001). 
Satoshi Sekine, Ralph Grishman and Hiroyuki 
Shinnou. 1998. A Decision Tree Method for 
Finding and Classifying Names in Japanese 
Texts.  Proc. Sixth Workshop on Very Large 
Corpora; Montreal, Canada. 
Jian Sun, Jianfeng Gao, Lei Zhang, Ming Zhou 
and Changning Huang. 2002. Chinese Named 
Entity Identification Using Class-based 
Language Model.  Coling 2002. 
Marc Vilain, John Burger, John Aberdeen, Dennis 
Connelly, Lynette Hirschman. 1995. A model --
Theoretic Coreference Scoring Scheme. MUC-6 
Proceedings, Nov. 1995. 
Shiren Ye, Tat-Seng Chua, Liu Jimin. 2002. An 
Agent-based Approach to Chinese Named 
Entity Recognition. Coling 2002. 
