This paper presents the results of our initial experiments in the monolingual English, Spanish and Portuguese tasks and the Bilingual Spanish ! English, Spanish ! Portuguese, English ! Spanish and Portuguese ! Spanish tasks. Twenty runs were submitted as o cial runs, thirteen for the monolingual task and seven for the bilingual task. We used the Terrier Information Retrieval Platform to run experiments for both tasks using the Inverse Document Frequency model with Laplace after-e ect and normalization 2. Experiments included topics processed automatically as well as topics processed manually. Manual processing of topics was carried out using gazetteers (Alexandria Digital Library, European Parliament and GEOnet Names Server), some of them containing translations in languages other than English, others containing the latitude, longitude and area which allow for semi-automated spatial analysis (proximity analysis). For the bilingual task we developed a component based on the transfer approach in machine translation. Topics were pre-processed automatically to eliminate stopwords. Then topics in the source language were translated to the target language. A major problem we detected after submitting our results was that we did not include the Spanish newspaper collection for the year 95 (EFE 95) for indexing and retrieval purposes. Therefore, the results of our experiments with Spanish for the monolingual and bilingual tasks were a ected in terms of recall and precision. We are currently rerunning experiments with the full Spanish collection for the monolingual and bilingual task to obtain a more accurate evaluation of the retrieval performance.
Geographic Information Retrieval (GIR) is aimed at the retrieval of geographic data based not only on conceptual keywords, but also on spatial information. Building GIR systems with such capabilities requires research on diverse areas such as information extraction of geographic terms from structured and unstructured data; word sense disambiguation, which is geographically relevant; ontology creation; combination of geographical and contextual relevance; and geographic term translation, among others.
Research e orts on GIR are addressing issues such as access to multilingual documents, techniques
for information mining (i.e., extraction, exploration and visualization of geo-referenced
information), investigation of spatial representations and ranking methods for di erent representations,
application of machine learning techniques for place name recognition, development of datasets
containing annotated geographic entities, among others. [
The purpose of GeoCLEF 2006 is to experiment with and evaluate the performance of GIR systems when topics includes geographic locations such as rivers, regions, seas, continents. Two tasks were considered, a monolingual and a bilingual. We participated in the monolingual task in English, Portuguese and Spanish; for the bilingual task we worked with topics in Spanish and documents in English and Portuguese, and with topics in English and Portuguese and documents in Spanish. In this paper we describe our initial experiments in the monolingual task and the bilingual task. We used the Terrier Information Retrieval (IR) platform to run our experiments, and built an independent module for the translation of the topics based on the transfer approach in machine translation. We used Terrier because it has performed successfully in monolingual information retrieval tasks in CLEF and TREC. Our goal is to have a baseline for further experiments with our component for translating georeferences and improved spatial analysis.
The paper is organized as follows. In Section 2 we present our work in the monolingual task including an overview of Terrier. Section 3 describes our setting and experiments in the bilingual task. We present conclusions and current work in Section 4. 2
In this section we rst give an overview of Terrier (TERabyte RetRIEveR) an information retrieval (IR) platform used in all the experiments. Then we describe the monolingual experiments for English, Portuguese and Spanish.
Terrier is a platform for the rapid development of large-scale Information Retrieval (IR) systems.
It o ers a variety of IR models based on the Divergence from Randomness (DFR) framework
([
The risk of accepting a term is inversely related to its term frequency in the document with respect to the elite set, a set in which the term occurs to a relatively greater extent than in the rest of the documents. The more the term occurs in the elite set, the less the term frequency is due to randomness. Hence the probability of the risk of a term not being informative is smaller. The Laplace model is utilized to compute the information gain with a term within a document. Term frequencies are calculated with respect to the standard document length using a formula referred to as normalization 2 shown below.
sl
tf n = tf:log(1 + c )
dl
tf is the term frequency, sl is the standard document length, and dl is the document length, c
is a parameter. We used c = 1.5 for short queries, which is the default value, c = 3.0 for short
queries with automatic query expansion and c = 5.0 for long queries. Short queries in our context
are those which use only the topic title and topic description; long queries are those which use
the topic title, topic description and topic narrative. We used these values based on the results
generated by the experiments on tuning for BM25 and DFR models done by He and Ounis [
We submitted 4 runs for English, 4 runs for Portuguese and 5 runs for Spanish. For some of the runs we used the automatic query expansion capability of terrier with the default values of 3 documents and 10 terms. We are running experiments with new values to determine which number of documents and terms perform better. Results for the monolingual task in English, Portuguese and Spanish are shown in Table 1, Table 2 and Table 3, respectively.
Construction
Expansion
Construction
Expansion title, description, narrative automatic yes title, description automatic yes title, description manual yes title, description automatic no title, description, narrative automatic no Table 3: Spanish Monolingual Retrieval Performance 14.71 14.71 14.71 13.78 14.71
title, description automatic yes title, description, narrative automatic yes title, description automatic no title, description, narrative automatic no Table 1: English Monolingual Retrieval Performance
For the bilingual task we worked with Spanish topics and English and Portuguese documents, and English and Portuguese topics and Spanish documents. We built a component, independent of Terrier, based on the transfer approach in machine translation to translate topics from the source language to the target language using mapping rules. All the information in the topics within the title, description and narrative was translated. Topics in English, Spanish, and Portuguese were preprocessed by removing diacritic marks and using stopwords lists. Diacritic marks were also removed from the stopwords lists and duplicates were eliminated. Plural stemming was then applied.
Automatic and manual query construction was carried out with the aid of the German Alexandria
Digital Library gazetteer [
Seven runs were submitted as o cial runs for the GeoCLEF2006 bilingual task. In Table 4 we report the results for X-Spanish (X=fEnglish, Portugueseg) and in Table 5 the results for SpanishX (X=fEnglish,Portugueseg). In this paper we presented work on monolingual and bilingual geographical information retrieval. We used Terrier to run our experiments, and an independent translation component built to map source language (English, Portuguese or Spanish) topics into target language (English, Portuguese or Spanish) topics. Results were a ected because we did not include one of the Spanish collections for indexing and retrieval purposes. We are currently re-running experiments with the entire collection of Spanish documents and testing the automatic query expansion capabilities of terrier with new values and new weighting models. [11] http://earth-info.nga.mil/gns/html/index.html