=Paper=
{{Paper
|id=Vol-1170/CLEF2004wn-adhoc-SavoyEt2004
|storemode=property
|title=Selection and Merging Strategies for Multilingual Information Retrieval
|pdfUrl=https://ceur-ws.org/Vol-1170/CLEF2004wn-adhoc-SavoyEt2004.pdf
|volume=Vol-1170
|dblpUrl=https://dblp.org/rec/conf/clef/SavoyB04
}}
==Selection and Merging Strategies for Multilingual Information Retrieval==
https://ceur-ws.org/Vol-1170/CLEF2004wn-adhoc-SavoyEt2004.pdf
Selection and Merging Strategies for Multilingual Information Retrieval
Jacques Savoy, Pierre-Yves Berger
Institut interfacultaire d'informatique
Université de Neuchâtel, Switzerland
{Jacques.Savoy, Pierre-Yves.Berger}@unine.ch Web site: www.unine.ch/info/clef/
Abstract. For our fourth participation in the CLEF evaluation campaigns, our objective was to
verify whether our combined query translation approach would work well with new requests and
new languages (Russian and Portuguese in this case). As a second objective, we suggested a
selecting procedure that could extract a smaller number of documents from collections that for the
current request seem to contain no or only few relevant items. We also applied different merging
strategies in order to obtain more evidence on the respective relative merits.
Introduction
Based on our bilingual and multilingual experiments of last years (Savoy 2003; 2004a), we conducted
different experiments involving various bilingual and multilingual test-collections. In the latter case, we
retrieved documents written in the English, French, Finnish and Russian languages, based on a request written
in English. As of last years, we adopted a combined query translation strategy that is able to produce queries in
different European languages based on an original request written in English. Once the translation phase was
completed, we searched in the corresponding document collection using our retrieval scheme (bilingual
retrieval). In Section 2, we carried out multilingual information retrieval, investigating various merging
strategies based on the results obtained during our bilingual searches.
1. Bilingual Information Retrieval
In our experiments, we chose English as the language to be used when submitting requests for automatic
translation into four different languages, using nine different machine translation (MT) systems and one
bilingual dictionary (“Babylon”). The following freely available translation tools were used in our experiments:
SYSTRAN www.systranlinks.com/
GOOGLE www.google.com/language_tools
FREETRANSLATION www.freetranslation.com/web.htm
INTERTRAN intertran.tranexp.com/
REVERSO www.reverso.fr/url_translation.asp
WORLDLINGO www.worldlingo.com/
BABELFISH babelfish.altavista.com/
PROMPT webtranslation.paralink.com/
ONLINE www.online-translator.com/srvurl.asp?lang=en
BABYLON www.babylon.com
When using the Babylon bilingual dictionary to translate an English request word-by-word, usually more
than one translation is provided, in an unspecified order. We decided to pick only the first translation available
(labeled “Babylon 1”), the first two terms (labeled “Babylon 2”) or the first three available translations (labeled
“Babylon 3”).
Table 1 shows the resulting mean average precision using the various translation tools and the Okapi
probabilistic model (see Savoy (2004c) for implementation details). Of course, not all tools can be used for
each language, and thus as shown in Table 1 various entries are missing (indicated with the label “N/A”). From
this data, we can see that the results from the FreeTranslation MT system usually obtain satisfactory retrieval
performances (around 82% of the MAP of the corresponding monolingual search). As another good translation
system, we may mention Reverso or BabelFish for the French, Prompt for the Russian or Online for both the
Russian and Portuguese languages. For the Finnish language we found only two translation tools, but
unfortunately their overall performance levels were not very good (a similar low level performance was also
found when translating English topics into various Asian languages (Savoy 2004b)). Not surprisingly, we
found there was a relationship between the various translation tools. For example, the Systran, BabelFish, and
WorldLingo MT systems appeared to be nearly identical MT systems.
Mean average precision (% of monolingual search)
Language French Finnish Finnish Russian Portuguese
Okapi Okapi Okapi Okapi Okapi
49 queries 45 queries 45 queries 34 queries 46 queries
Manual 0.4685 0.4773 0.5386 0.3800 0.4835
Systran 0.3729 (79.6%) N/A N/A 0.2077 (54.7%) 0.3329 (68.9%)
Google 0.3680 (78.5%) N/A N/A N/A 0.3375 (69.8%)
FreeTrans 0.3845 (82.1%) N/A N/A 0.3067 (80.7%) 0.4057 (83.9%)
InterTrans 0.2664 (56.9%) 0.2290 (48.0%) 0.2653 (49.3%) 0.1216 (32.0%) 0.3277 (67.8%)
Reverso 0.3830 (81.8%) N/A N/A N/A N/A
WorldLingo 0.3728 (79.6%) N/A N/A 0.2077 (54.7%) 0.3311 (68.5%)
BabelFish 0.3729 (79.6%) N/A N/A 0.2077 (54.7%) 0.3329 (68.9%)
Prompt N/A N/A N/A 0.2960 (77.9%) N/A
Online N/A N/A N/A 0.2888 (76.0%) 0.3879 (80.2%)
Babylon 1 0.3706 (79.1%) 0.1771 (37.1%) 0.1965 (36.5%) 0.2209 (58.1%) 0.3071 (63.5%)
Babylon 2 0.3356 (71.6%) N/A N/A 0.2245 (59.1%) 0.2892 (59.8%)
Babylon 3 0.3378 (72.1%) N/A N/A 0.2243 (59.0%) 0.2858 (59.1%)
Table 1: Mean average precision of various single translation devices (TD queries, Okapi model)
It is known that while a given translation tool may produce acceptable translations for a given set of
requests, it may perform poorly for other queries (Savoy 2003; 2004a). To date we have not been able to detect
very precisely when a given translation will produce satisfactory retrieval performance and when it will fail. In
this vein, Kishida et al. (2004) suggest using a linear regression model to predict the average precision of the
current query, based on both manual evaluations of translation quality for the current query and the underlying
topic difficulty. In this study, before carrying out the retrievals, we chose to concatenate two or more transla-
tions before submitting a query for translation.
Mean average precision
Language French Finnish Finnish Russian Portuguese
Okapi Okapi Okapi Okapi Okapi
Combination 49 queries 45 queries 45 queries 34 queries 46 queries
Comb!1 Bab2+Free Bab1+Inter Bab1+Inter Bab1+Free Free+Online
Comb!2 Bab2+Reverso Free+Prompt Bab1+Systran
Comb!3 Reverso+Systran Prompt+Online Bab1+Free+Onl
Comb!4 Free+Rev Free+Online Bab1+Free+Sys
Comb!5 Bab2+Free+ Bab1+Free+ Bab1+Free+
Reverso Online Online+Systran
Best single 0.3845 0.2290 0.2653 0.3067 0.4057
Comb!1 0.3784 0.2529 0.3042 0.3888 0.4072
Comb!2 0.3857 0.3032 0.3713
Comb!3 0.3858 0.2964 0.4204
Comb!4 0.4066 0.3043 0.3996
Comb!5 0.3962 0.3324 0.4070
Table 2: Mean average precision of various combined translation devices (TD queries, Okapi model)
Table 2 shows the retrieval effectiveness for such combinations, using the Okapi probabilistic model. The
top part of the table indicates the exact query translation combination used while the bottom part shows the
mean average precision achieved by our combined query translation approach. When selecting the query
translations to be combined, a priori we considered the best translation tools.
The resulting retrieval performances shown in Table 2 are sometimes better than the best single translation
scheme indicated in the row labeled “Best single” (e.g., the strategies “Comb 4” or “Comb 5” for French, or
“Comb 1” for Russian, and “Comb 3” for the Portuguese language). Of course, the main difficulty in this
bilingual search was the translation of English topics into Finnish, due to limited number of free translation
tools. When handling those languages less-often speaking around the world, it seems it would be worthwhile
considering other translation alternatives, such as probabilistic translation based on parallel corpora (Nie et al.
1999), (MacNamee & Mayfield 2003).
For monolingual searches, as described in Savoy (2004c), we used a data fusion search strategy that
combined the Okapi and Prosit probabilistic models (see details in Section 2). The data shown in Table 3
indicates that our data fusion approaches may result in better retrieval effectiveness (except for the Finnish 4-
gram indexing scheme or the Russian corpus). Of course before combining the result lists we could also
automatically expand the translated queries, using a pseudo-relevance feedback method (Rocchio’s approach in
the present case). The resulting mean average precision as shown in Table 4 did not improve the retrieval
effectiveness when compared to the best single approach. In Tables!3 and 4, under the heading “Z-scoreW”, we
attached a weight of 1.5 to the Prosit model, and 1 to the Okapi model. Finally, Table 5 depicts the parameters
used for our official bilingual runs.
Mean average precision
Language French Finnish Russian Portuguese
word 4-gram word word
49 queries 45 queries 34 queries 46 queries
Combination \ Translation Comb!5 Comb!1 Comb!4 Comb!3
Okapi 0.3962 0.3042 0.3043 0.4204
Prosit 0.3937 0.2853 0.2928 0.4085
Round-robin 0.3950 0.2969 0.2943 0.4129
Sum!RSV 0.3980 0.2965 0.3036 0.4134
Norm!Max 0.3977 0.2935 0.3010 0.4152
Norm!RSV (Eq.1) 0.3978 0.2937 0.3010 0.4152
Z-score 0.3980 0.2937 0.3014 0.4152
Z-scoreW 0.3973 0.2965 0.3009 0.4043
Table 3: Mean average precision of automatically translated queries (without automatic query expansion)
Mean average precision
Language French Finnish Russian Portuguese
word 4-gram word word
49 queries 45 queries 34 queries 46 queries
Combination \ Translation Comb!5 Comb!1 Comb!4 Comb!3
Okapi (#docs/#terms) 0.4071 (5/15) 0.2956 (5/30) 0.3110 (3/15) 0.4315 (5/15)
Prosit (#docs/#terms) 0.4055 (5/10) 0.2909 (10/30) 0.2914 (3/15) 0.4724 (10/20)
Round-robin 0.4153 0.2999 0.3007 0.4637
Sum!RSV 0.4096 0.2928 0.3000 0.4611
Norm!Max 0.4091 0.2964 0.3070 0.4711
Norm!RSV (Eq.1) 0.4126 0.2967 0.3086 0.4704
Z-score 0.4118 0.2955 0.3073 0.4699
Z-scoreW 0.4098 0.2948 0.3024 0.4722
Table 4: Mean average precision of automatically translated queries (after blind query expansion)
Russian Russian Portuguese Portuguese
34 queries 34 queries 46 queries 46 queries
IR model 1 (#docs/#terms) Prosit (3/15) Prosit (3/15) Prosit (10/20) Okapi (0/0)
IR model 2 (#docs/#terms) Okapi (3/15) Okapi (3/10) Okapi (5/15) Prosit (0/0)
Data fusion operator Round-robin Round-robin Norm RSV Norm RSV
Translation tools Free-Reverso Pro-Free-Reverso Onl-Free-Bab1 Onl-Free-Sys-Bab1
Mean average precision 0.3007 0.2962 0.4704 0.4491
Run name UniNEBru1 UniNEBru2 UniNEBpt1 UniNEBpt2
Table 5: Description and mean average precision (MAP) of our official bilingual runs
2. Multilingual Information Retrieval
Our multilingual information retrieval system is based on the use of a query translation strategy instead of
either translating all documents into a common language (e.g., English), combining both query and document
translations (Chen & Gey 2003) or ignoring the translation phase (Buckley et al. 1998), (MacNamee &
Mayfield 2003); for a general overview of these questions, see (Braschler & Peters 2004). In our approach,
when a request was received (in English in this study), we automatically translated it into the desired target
languages and then searched for pertinent items within each of the four corpora (English, French, Finnish and
Russian). After receiving a result list from each search engine, we needed to introduce a merging procedure to
provide a unique ranked result list. As a first approach to this problem, we considered the round-robin approach
whereby we took one document in turn from each individual list (Voorhees et al. 1995).
To account for the document score computed for each retrieved item (denoted RSVk for document Dk), we
might formulate the hypothesis that each collection is searched by the same or a very similar search engine and
that the similarity values are therefore directly comparable (Kwok et al. 1995). Such a strategy is called raw-
score merging and produces a final list sorted by the document score computed by each collection. When using
the same IR model (with the same or very similar parameter settings) to search into all collections, such a
merging strategy may produce good retrieval performance (e.g., with a logistic regression IR model in (Chen
2003)).
Unfortunately the document scores cannot always be directly compared, thus as a third merging strategy we
normalized the document scores within each collection by dividing them by the maximum score (i.e. the
document score of the retrieved record in the first position) and denoted them “Norm Max”. As a variant of this
normalized score merging scheme (denoted “Norm RSV”), we could normalize the document RSVk scores within
the ith result list, according to the following formula:
Norm RSVk = ((RSVk - MinRSVi ) / (MaxRSVi - MinRSVi )) (1)
As a fifth merging strategy, we might use logistic regression to predict the probability of a binary outcome
variable, according to a set of explanatory variables (Le Calvé & Savoy 2000). In our current case, we predicted
the probability of relevance of document Dk given both the logarithm of its rank (indicated by ln(rankk)) and the
original document score RSVk as indicated in Equation 2. Based on these estimated relevance probabilities
(computed independently for each language using the S+ software), we sorted the records retrieved from separate
collections in order to obtain a single ranked list. However, in order to estimate the underlying parameters, this
approach requires that a training set is available. To achieve this, we used the CLEF-2003 topics and their
relevance assessments in our evaluations.
ea+b1 ⋅ln(rank k )+b 2 ⋅rsv k
Pr ob [ D k is rel | rank k , rsv k ] = (2)
1 + ea+b1 ⋅ln(rank k )+b 2 ⋅rsv k
Parameters of each single run according to each language
TD Queries English French Finnish (4-gram) Russian (word)
† 42 queries 49 queries 45 queries 34 queries
Condition A
IR model 1 (#docs/#terms) Okapi (3/15) Prosit (5/15) Okapi (5/30) Prosit (3/15)
IR model 2 (#docs/#terms) Prosit (3/10) Okapi (5/10)
Data fusion operator Z-score Z-scoreW
Translation tools Bab2-Free-Rev Bab1-Inter Rev-Free
Mean average precision 0.5580 0.4098 0.2956 0.2914
Condition B
IR model 1 (#docs/#terms) Okapi (3/15) Prosit (5/15) Okapi (5/30) Prosit (3/15)
IR model 2 (#docs/#terms) Prosit (3/10) Okapi (5/10) Lnu-ltc (3/40) Okapi (3/15)
Data fusion operator Z-score Z-scoreW Round-robin Round-robin
Translation tools Bab2-Free-Rev Bab1-Inter Rev-Free
Mean average precision 0.5580 0.4098 0.3080 0.3007
Condition C
IR model (#docs/#terms) Prosit (3/10) Prosit (5/15) Prosit (10/30) Prosit (3/15)
Translation tools Bab2-Fre-Rev Bab1-Inter Rev-Free
Mean average precision 0.5633 0.4055 0.2909 0.2914
Table 6: Description of the various runs done separately on each corpus (top descriptions form the
Condition A, middle descriptions form Condition B, and bottom descriptions form Condition C)
Finally, we suggest merging the retrieved documents according to the Z-score, taken from their document
scores (Savoy 2003). Within this scheme, we need to compute, for the ith result list, the average of the RSVk
(denoted MeanRSVi ) and the standard deviation (denoted StdevRSVi ). Based on these values, we can normalize
the retrieval status value of each document Dk provided by the ith result list, by computing the following
formula:
Z-Score RSVk = a i . [((RSVk-MeanRSVi ) / StdevRSVi ) + d i] with di = ((MeanRSVi - MinRSVi )/StdevRSVi ) (3)
within which the value of di is used to generate only positive values, and a i (usually fixed at 1) is used to
reflect the retrieval performance of the underlying retrieval model and to account for the fact that pertinent items
are not uniformly distributed across all collections.
Table 6 depicts the exact parameters used to search in the four different collections. For the Russian
collection, we only considered the word-based indexing strategy while for the Finnish language we only used
the 4-gram indexing scheme. In the top part of Table 6, it can be seen that we used a combined query
translation strategy for French, Finnish and Russian languages. As described in our monolingual experiments
(Savoy 2004c), we might also apply a data fusion phase before merging the result lists. Thus when searching
into the English or French corpus, we combined the Okapi and Prosit result lists (both with blind query
expansion). In a second multilingual experiment (denoted Condition B), we have applied a data fusion approach
for all bilingual searches (descriptions given in the middle part of Table 6). Finally, we decided to search
through all corpora using the same retrieval model, Prosit in this case, as shown in the bottom part of Table 6
(and corresponding to Condition C).
Table 7 depicts the retrieval effectiveness of various merging strategies using three different bilingual search
parameter settings. In this table, the round-robin scheme will be used as a baseline. On the one hand, when
different search engines are merged (Condition A and Condition B), the raw-score merging strategy results in
very poor mean average precision. On the other hand, when the same search engine is used (Condition C), the
resulting performance is better, but this is not the best one we should be able to achieve. The normalized score
merging based on Equation 1 shows degradation over the simple round-robin approach when using parameter
setting Condition B (0.1042 vs. 0.2340, or -4.9% in relative performance). Applying our logistic model using
both the rank and the document score as explanatory variables, the resulting mean average precision is clearly
better than the round-robin merging strategy and than other merging approaches (under Condition A or C).
Under Condition B, the difference between our logistic model and the Z-score merging strategy is rather small
(0.3111 vs. 0.3019, or 3.1% in relative performance).
As a simple alternative, we also suggest a biased round-robin approach which extracts not one document per
collection per round but one document for the Russian corpus and two from the English, French and Finnish
collection (because the last three represent larger corpora). This merging strategy results in good retrieval
performance, better that the simple round-robin approach. Finally, the Z-score merging approach seems to
provide generally satisfactory performance. Moreover, we may multiply the Z-score by an a value (performance
under the label “ai = 1.5” with the a i values set as follows: EN: 1.5, FR: 1.5, FI: 1.0, and RU: 1.0).
Mean average precision (% change)
Parameters setting Condition A Condition B Condition C
Merging Strategy 50 queries 50 queries 50 queries
Round-robin (baseline) 0.2386 0.2430 0.2358
Raw-score 0.0642 (-73.1%) 0.0650 (-73.2%) 0.3067 (+30.1%)
Norm Max 0.2552 (+7.0%) 0.1044 (-57.0%) 0.2484 (+5.3%)
Norm RSV (Eq. 1) 0.2899 (+21.5%) 0.1042 (-57.1%) 0.2646 (+12.2%)
Logistic reg. (ln(rank), RSV) 0.3090 (+29.5%) 0.3111 (+28.0%) 0.3393 (+43.9%)
Biased round-robin 0.2639 (+10.6%) 0.2683 (+10.4%) 0.2613 (+10.8%)
Z-score (Eq. 3) 0.2677 (+12.2%) 0.2903 (+19.5%) 0.2555 (+8.4%)
Z-score (Eq. 3) ai = 1.5 0.2669 (+11.9%) 0.3019 (+24.2%) 0.2867 (+21.6%)
Logistic reg. & Selection (0) 0.2957 (+23.9%) 0.2959 (+21.8%) 0.3405 (+44.4%)
Logistic reg. & Selection (3) 0.2953 (+23.8%) 0.2982 (+22.7%) 0.3378 (+43.3%)
Logistic reg. & Selection (10) 0.2990 (+25.3%) 0.3008 (+23.8%) 0.3381 (+43.4%)
Logistic reg. & Selection (20) 0.3010 (+26.1%) 0.3029 (+24.7%) 0.3384 (+43.5%)
Logistic reg. & Selection (50) 0.3044 (+27.6%) 0.3064 (+26.1%) 0.3388 (+43.7%)
Logistic reg. & OptimalSelect 0.3234 (+35.5%) 0.3261 (+34.2%) 0.3558 (+50.9%)
Table 7: Mean average precision of various merging strategies (TD queries)
It cannot be expected however that each result list would always contain pertinent items in response to a
given request. In fact, a given corpus may contain no relevant information regarding the submitted request or
the pertinent articles cannot be found by the search engine. In a cross-lingual environment we have found an
additional problem: important facets of the original request were translated with inappropriate words or
expressions. In all these cases, it is not useful to include items provided by such collections (or such search
engines) in the final result list. In addition, the number of pertinent documents is usually not uniformly
distributed across all four collections. For a given request (e.g., related to a regional or a national event), only
one or two collections may contain relevant documents describing this particular event.
To take into account these phenomena, we have designed a selection procedure which works as follows.
First, for each result list we normalize the document score according to our logistic regression method (given in
Equation 2). After this step, each document score represents the probability that the underlying article is
relevant (with respect to the submitted query and the collection). In the second step, for each result list (or
language) we sum the document scores of the first 15 top-ranked documents. If this sum exceeds a given
threshold (depending on the collection or search engine), we can thus consider that the corresponding collection
contains many pertinent documents. Otherwise, we might only include the m best ranking retrieved items from
the corpus (with a relatively small m value). We may thus limit the number of items extracted from a given
corpus while also taking account of the fact that each collection usually contains few pertinent items. Table 7
lists the mean average precision achieved using this selection strategy under the label “Logistic reg. & Selection
(m),” where the value m indicates that we always include the m best retrieved items from each corpus in our final
result list. Of course, when we set m = 0, the system will not extract any documents from a collection having a
poor overall score. Finally under the label “Logistic reg. & OptimalSelect“, we have computed the mean
average precision that can be achieved when the selection is done without any error (with m = 0). When using
such an ideal selection system, the mean average precision is clearly better than all other merging strategies (e.g,
under Condition C, the MAP is 0.3558 vs. 0.3393 with the logistic regression without selection).
Table 8 contains the descriptions of our official runs for the multilingual tracks. In the row entitled
UniNEmulti3, all searches were done based on the Prosit retrieval model in order to obtain more comparable
document score across the various collections.
Run name Query lang. Query type Type Merging Parameters MAP
UniNEmulti1 English TD automatic logistic Condition A 0.3090
UniNEmulti2 English TD automatic Z-scoreW Cond.!A, a i !=!1.5 0.2969
UniNEmulti3 English TD automatic raw-score Condition C 0.3067
UniNEmulti4 English TD automatic logistic & select Cond. A, m = 20 0.3010
UniNEmulti5 English TD automatic Z-scoreW Condition B 0.3019
Table 8: Description and mean average precision (MAP) of our official multilingual runs
Conclusion
In this fifth CLEF evaluation campaign, we evaluated various query translation tools (see Table 1), together
with a combined translation strategy (see Table 2), resulting in a retrieval performance that is worth considering.
However, while a bilingual search can be viewed as easier for some language pairs (e.g., from an English query
into a French document collection, or English to Portuguese), this task is clearly more complex for other
language pairs (e.g., English to Finnish). Combining different result lists (see Table 3 or 4), we cannot always
obtain a better retrieval effectiveness compared to isolated runs.
In multilingual tasks, searching documents written in different languages represents a real challenge. In this
case we propose a new simple selecting strategy which will avoid extracting a relatively large number of
documents from collections when these documents are of little interest with respect to the current request (see
Table 7). In this multilingual task, it is also interesting to mention that combining the result lists provided by
same search engine (Condition C in Table 7) may sometimes produce good retrieval effectiveness compared to
combining different search models (Condition A in Table 7).
Acknowledgments
The authors would like to thank the CLEF-2004 task organizers for their efforts in developing various
European languages test-collections. The authors would also like to thank C. Buckley from SabIR for giving us
the opportunity to use the SMART system, together with Samir Abdou for his help in translating the English
topics. This research was supported by the Swiss National Science Foundation under Grant #21-66 742.01.
References
Braschler, M. & Peters, C. (2004). Cross-language evaluation forum: Objectives, results and achievements. IR
Journal, 7(1-2), 7-31.
Buckley, C., Mitra, M., Waltz, J. and Cardie, C. (1998). Using clustering and superconcepts within SMART.
In Proceedings of TREC-6, (pp. 107-124). Gaithersburg: NIST Special Publication 500-240.
Chen, A. & Gey, F. (2003). Combining query translation and document translation in cross-language retrieval.
In Proceedings CLEF-2003, (pp. 39-48). Trondheim.
Chen, A. (2003). Cross-language retrieval experiments at CLEF 2002. In C. Peters, M. Braschler, J. Gonzalo,
& M. Kluck, (Eds), Advances in Cross-Language Information Retrieval, (pp. 28-48), Springer-Verlag,
Berlin, LNCS #2785.
Kishida, K., Kuriyama, K., Kando, N. & Eguchi, K. 2004. Prediction of performance on cross-lingual
information retrieval by regression models. In Proceedings NTCIR-4, (pp. 219-224). Tokyo: NII.
Kwok, K.L., Grunfeld, L. & Lewis, D.D. (1995). TREC -3 ad-hoc, routing retrieval and thresholding
experiments using PIRCS. In Proceedings of TREC’3, (pp. 247-255). Gaithersburg: NIST Publication #500-
225.
Le Calvé, A. & Savoy, J. (2000). Database merging strategy based on logistic regression. Information
Processing & Management, 36(3), 341-359.
MacNamee, P. & Mayfield, J. (2003). JHU/APL experiments in tokenization and non-word translation. In
Proceedings CLEF-2003, (pp. 19-28). Trondheim.
Nie, J. Y., Simard, M., Isabelle, P. & Durand, R. (1999). Cross-language information retrieval based on
parallel texts and automatic mining of parallel texts from the Web. In Proceedings of the ACM-SIGIR'99,
(pp. 74-81). New York: The ACM Press.
Savoy J. (2003). Report on CLEF-2003 multilingual tracks. In Proceedings CLEF-2003, (pp. 7-12).
Trondheim.
Savoy, J. (2004a). Combining multiple strategies for effective monolingual and cross-lingual retrieval. IR
Journal, 7(1-2), 121-148.
Savoy, J. (2004b). Report on CLIR task for the NTCIR-4 evaluation campaign. In Proceedings NTCIR-4, (pp.
178-185). Tokyo: NII.
Savoy, J. (2004c). Report on CLEF-2004 monolingual tracks. In Proceedings CLEF-2004 (this volume). Bath.
Voorhees, E.M., Gupta, N.K. & Johnson-Laird, B. (1995). The collection fusion problem. In Proceedings of
TREC’3, (pp. 95-104). Gaithersburg: NIST Publication #500-225.
C201 Domestic Fires C226 Sex-change Operations
C202 Nick Leeson's Arrest C227 Altai Ice Maiden
C203 East Timor Guerrillas C228 Prehistorical Art
C204 Victims of Avalanches C229 Dam Building
C205 Tamil Suicide Attacks C230 Atlantis-Mir Docking
C206 G7 Summit in Halifax C231 New Portuguese Prime Minister
C207 Fireworks Injuries C232 Pension Schemes in Europe
C208 “Sophie's World” C233 Greenhouse Effect
C209 Tour de France Winner C234 Deaf and Society
C210 Nobel Peace Prize Candidates C235 Seal-hunting
C211 Peru-Ecuador Border Conflict C236 A typhoon in the Philippines
C212 Sportswomen and Doping C237 Panchen Lama
C213 Papal Travels C238 Lady Diana
C214 Multi-billionaires C239 Mental Health of the Young
C215 Re-election of Peru's President C240 Sioux Ghost Shirt
C216 Glue-sniffing Youngsters C241 New political parties
C217 AIDS in Africa C242 Record Permanence in Space
C218 Andreotti and the Mafia C243 Films of Kieslowski
C219 EU Commissioner Candidates C244 Footballer of the Year 1994
C220 European Cars in Russia C245 Christopher Reeve
C221 2002 Olympic Winter Games C246 Castro visits UN
C222 Presidential elections in France C247 Alexander the Great's Tomb
C223 Chernobyl Disaster outside ex-USSR C248 Macedonia Name Dispute
C224 Woman solos Everest C249 Women's Ten Thousand Metres Champion
C225 Nuclear Power Plant of Sosnovyi Bor C250 Rabies in Humans
Title of the queries of the CLEF-2004 test-collection