ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001 
204 
Relevance 
results 
Irrelevance 
results 
No-Name 
1951 
547 
Yahoo 
1593 
2640 
AlltheWeb 
1145 
3256 
Geo-Community 
125 
23 
Table 3 : Query results of term set with 50 terms 
Natural Language Support and Linguistic Analysis 
Two sentences, "Tell me all about GPS activities in China" and 
"I want to know something about AM/FM applications in 
Canada", are selected as queries to test the ability of No- 
Name and current search engines to support natural language 
input and linguistic analysis. The results are shown in Table 4 
and Table 5. While "GPS" stands for "Global Positioning 
Satellite System", "AM/FM" stands for "Automated Mapping 
and Facilities Management". 
Relevance 
results 
Irrelevance 
results 
No-Name 
261 
43 
Yahoo 
132 
68 
AlltheWeb 
1 
199 
Geo-Community 
0 
0 
Table 4 : Query results of the first sentence 
Relevance 
results 
Irrelevance 
results 
No-Name 
42 
7 
Yahoo 
10 
190 
AlltheWeb 
0 
200 
Geo-Community 
0 
11 
Table 5 : Query results of the second sentence 
Comparison of the system performances before learning 
and after learning 
No-Name can revise the term connection value by learning 
from the users' feedback. A category name "Database" is 
chosen as termX, and 5 terms, ("Binary Large Object", 
"Conceptual Model", "Data Definition Language", 
"Georelational Model", and "Spatial Database") in this category 
are chosen to represent the queries. The initial term 
connection values between termX and the 5 terms are set to 
0.5. Each query can only contain one term. The results of the 
evaluation from expert board are used as users' feedback to 
train the search engine. 100 results are chosen as feedback 
from the total 408 evaluation results randomly. The query 
results before learning and after learning are shown in Table 6 
and the term connection values before learning and after 
learning are shown in Table 7. 
No-Name 
Relevance 
results 
Irrelevance 
results 
Before learning 
271 
137 
After learning 
256 
92 
Table 6 : Query results before learning and after 
learning 
Term 
Connection 
value 
TermX 
(before 
learning) 
TermX 
(after 
learning) 
Termi 
0.5 
0.7 
Term2 
0.5 
0.3 
Term3 
0.5 
0.8 
Term4 
0.5 
0.5 
Term 5 
0.5 
0.9 
Table 7 : Term connection values before learning 
and after learning 
CONCLUSIONS 
Compared to current search engines, No-Name, an Intelligent 
GIS search engine, has a higher performance in Relevant 
Sites Retrieval, Irrelevant Sites Dodge, and Natural Language 
Identification. 
Spider, robot, and full text analysis technology are applied in 
No-Name successfully. The information collecting from Internet 
is efficient and adequate. 
It is realized to determine the strength of term connections in 
small size thesauri according to users' feedback using fuzzy 
logic. The experiment results support the search engine can 
learn from the users’ feedback. The relevance results are 
almost kept as same, but the irrelevance results are highly 
reduced after training. 
The test results show No-Name has the best performance in 
natural language supporting among the four chosen search 
engines. Anyway, Yahoo also shows now it can support 
natural language identification somehow. The improvement 
should be attributed to the partnership between Yahoo and 
Google that began in June of 2000 [27]. 
No-Name's unique multiple independent layers structure can 
speed the query and help the users to find better matched 
results. 
Although the experimental results shows No-Name is an 
Intelligent GIS search engine to retrieve GIS information from 
Internet with high performance, there are still some room left 
for improvement. First, the expert board should contain more 
experts to avoid misjudgment. The decisions made by only a 
few experts are usually fatal to misjudgment and thus become 
doubtful. The noise, or misjudgment by one expert is expected 
to be weaken by recruiting more experts to the expert board. 
Second, The assumption 2, that any term has no connection 
with its peers in the same category, is too strong and might be 
not the truth. The term connection value among terms in the 
same category should be determined in the future which 
means the number of term connection will be increased from a 
few thousand to several tens of thousand. Finally, GIS is an 
explosive interdisciplinary application science, and the terms 
that GIS involves with are increasing dramatically. It is a 
challenging problem to maintain a reasonable size GIS 
thesauri for the efficiency's sake, and at the same time, to 
keep up with the advancement of GIS technologies.