=Paper=
{{Paper
|id=Vol-1173/CLEF2007wn-GeoCLEF-NasikhinEt2007
|storemode=property
|title=Location Identification for Geographic Information Retrieval
|pdfUrl=https://ceur-ws.org/Vol-1173/CLEF2007wn-GeoCLEF-NasikhinEt2007.pdf
|volume=Vol-1173
|dblpUrl=https://dblp.org/rec/conf/clef/NasikhinA07
}}
==Location Identification for Geographic Information Retrieval==
https://ceur-ws.org/Vol-1173/CLEF2007wn-GeoCLEF-NasikhinEt2007.pdf
Location Identification for the Geographic information
Retrieval
Nasikhin and Mirna Adriani
Faculty of Computer Science
University of Indonesia
Depok 16424, Indonesia
{nasikhin, mirna}@cs.ui.ac.id
Abstract. In this paper we identify location names that appear in queries written
in Indonesian using geographic gazeeter. We built the gazeeter by collecting
geographic information from a number of geographic resources. We translated
an Indonesian query set into English using a machine translation technique. We
also made an attempt to improve the retrieval effectiveness using a query
expansion technique. The result shows that identifying locations in the queries
and applying the query expansion technique can help improve the retrieval
effectiveness for certain queries.
Keywords: cross-language information retrieval, geographic information retrieval, query
expansion.
1 Introduction
As our participation in the Geographical Information Retrieval of the Cross Language
Evaluation Forum (CLEF 2007) task, i.e., for Indonesian-English, we needed to use
language resources to translate Indonesian queries into English. We learned from our
previous work [1, 2] that freely available dictionaries on the Internet could not
correctly translate many Indonesian terms, as their vocabulary was very limited.
Luckily we found a machine translation tool available on the Internet that could help
translate the Indonesian queries into English. However, GIR focuses on identifying
geographical names [4] that appear in the queries, so we also needed to work on
translating the location names from English to Indonesian.
2 The Process of Identifying Location Names
There are many resources that contain geographical information available on the
Internet. We made use of the Gazeeters to build a location hierarchy map. The
location hierarchy was built by extracting the names of countries, province, etc. from
the Gazeeters. For each location, information about other locations within the area
that it covers was added, such as cities under a province, etc. We obtained the needed
geographical information from the Geonames (http://www.geonames.org/) and
Wikipedia (http://id.wikipedia.org).
We extracted the names of provinces, their capital cities, the names of mountains, seas
etc. from the Geonames and its translation in Bahasa Indonesia from the Wikipedia.
Each location has all their alternate names in both English and Bahasa Indonesia. If
one location name appears in a query or document, it will be looked up in the
gazeeters to find its associated locations that can be used as terms for searching for or
indexing the document.
Most documents in the collection contain information about the location of events.
For each document, we identified the location where the event mentioned in the
document occurred, and added the location information into the document’s index
entry. For documents that contain more than one location, we choose the location that
has the highest frequency in the document. If there is more than one location with the
same highest frequency then a location is selected randomly among such locations.
To process the geographical locations further, we identify words that are related to a
location name such as in the (north/south/…) of, in the border of, around etc. Then we
include all location names that fall inside a boxline surrounding the location (city,
country etc.) The boxline borders are at certain distance north, south, east, and west of
a location.
2.1 Query Expansion Technique
Adding translated queries with relevant terms (query expansion) has been shown to
improve CLIR effectiveness [1, 3]. One of the query expansion techniques is called
the pseudo relevance feedback [5]. This technique is based on an assumption that the
top few documents initially retrieved are indeed relevant to the query, and so they
must contain other terms that are also relevant to the query. The query expansion
technique adds such terms into the previous query. We applied this technique in this
work. To choose the relevant terms from the top ranked documents, we used the tf*idf
term weighting formula [5]. We added a certain number of terms that have the highest
weight scores.
3 Experiment
We participated in the bilingual task with English topics. The English document
collection contains 190,604 documents from two English newspapers, the Glasgow
Herald and the Los Angeles Times. We opted to use the query title and the query
description that came with the query topics. The query translation process was
performed fully automatic using a machine translation technique. The machine
translation technique translates the Indonesian queries into English using Toggletext1,
a machine translation tool that is available on the Internet.
We then applied a pseudo relevance-feedback query-expansion technique to the
queries that were translated using three techniques above. Beside adding terms, we
also add location names only that appear on the top documents. In these experiments,
we used Lemur2 information retrieval system which is based on the language model to
index and retrieve the documents.
4 Results
Our work focused on the bilingual task using Indonesian queries to retrieve
documents in the English collections. Table 1 shows the result of our experiments.
Table 1. Average retrieval precision of the monolingual runs of the title, their translation
queries, and the use of the geographic identification and query expansion on the translated
queries.
Task Monolingual % Change
Title 0.1767 -
Title (translation) 0.1417 -19.80%
Title (Geoprocessing) 0.1736 -1.75%
Title (Geoprocessing + Geofeedback) 0.1389 -21.39%
Title (Geoprocessing + Pseudofeedback) 0.1936 +9.56%
The retrieval performance of the title-based translation queries dropped 19.80% below
that of the equivalent monolingual retrieval (see Table 1). The retrieval performance
of location identification process on the queries dropped 1.75% below that of the
equivalent monolingual queries. Expanding the queries by adding geographic location
from the top documents to the translated queries decreases the retrieval performance
by 21.39%. However, adding terms that appear on the top documents on the translated
queries improve the retrieval performance by 9.56%.
The retrieval performance of the combination of title and description queries that is
translated by machine translation dropped 8.43% below that of the equivalent
monolingual retrieval (see Table 2). The identification of the location on the queries
improves the average precision 5.91%. Expanding the queries by adding the
geographic location that appears from the top documents increase the average
precision by 5.65%. However, adding terms from the top documents decrease the
average precision by 2.02%.
1
See http://www.toggletext.com/.
2
See http://www.lemurproject.org/.
Table 2. Average retrieval precision of the monolingual runs of the combination of title and
description topics, their translation queries, and the use of the geographic identification and
query expansion on the translated queries.
Task Monolingual % Change
Title + Description 0.1979 -
Title + Description (translation) 0.1812 -8.43%
Title + Description (Geoprocessing) 0.2096 +5.91%
Title + Description (Geoprocess + Geofeedback) 0.2091 +5.65%
Title + Description (Geoprocess + Pseudofeedback) 0.1939 -2.02%
5 Summary
Our results demonstrate that identifying location on the queries can have positive and
negative effect on the queries. The query expansion technique that was applied to the
queries by adding more terms and location names also produced mixed results. For
the title queries, the query expansion had a positive impact when the combination of
terms and location names were added to the queries. However, the same situation did
not work for the combination of title and description queries. It had a positive impact
only when the queries were added with terms or location names only. We still need to
study further on the effect of location identification because the decreased in retrieval
performance was not only caused by the failure in identifying the correct location
names but also the failure in translating the words and location names in the queries
from one language to another language.
References
1. Adriani, M. and C.J. van Rijsbergen. Term Similarity Based Query Expansion for Cross
Language Information Retrieval. In Proceedings of Research and Advanced Technology
for Digital Libraries, Third European Conference (ECDL’99), p. 311-322. Springer
Verlag: Paris, September 1999.
2. Buscaldi, D., Rosso, P., Garcia, P. P. WordNet-based Index Terms Expansion for
Geographical Information Retrieval. In the In Working Notes for the CLEF 2006
Workshop, 2006.
3. Larson, Ray. Cheshire II at GeoCLEF: Fusion and Query Expansion for GIR. . In
Proceedings of the Geographic Information Retrieval CLEF (GeoCLEF’05). Vienna,
Austria, 2005.
4. Overell, S. E., Ruger, S. Identifying and Grounding Descriptions of Places. In the
Proceedings of the Geographic Information Retrieval workshop (GIR’06). Seattle, USA,
2006.
5. Salton, Gerard, and McGill, Michael J. Introduction to Modern Information Retrieval,
New York: McGraw-Hill, 1983.