=Paper=
{{Paper
|id=Vol-1169/CLEF2003wn-adhoc-QuEt2003
|storemode=property
|title=Clairvoyance CLEF-2003 Experiments
|pdfUrl=https://ceur-ws.org/Vol-1169/CLEF2003wn-adhoc-QuEt2003.pdf
|volume=Vol-1169
|dblpUrl=https://dblp.org/rec/conf/clef/QuGE03a
}}
==Clairvoyance CLEF-2003 Experiments==
https://ceur-ws.org/Vol-1169/CLEF2003wn-adhoc-QuEt2003.pdf
Clairvoyance CLEF-2003 Experiments
Yan Qu, Greg Grefenstette, David A. Evans
Clairvoyance Corporation
5001 Baum Boulevard, Suite 700
Pittsburgh, PA 15213-1854
{yqu, grefen, dae}@clairvoyancecorp.com
Abstract
In CLEF 2003, Clairvoyance participated in the bilingual retrieval track with the German and Italian language
pair. As we did not have any German-to-Italian translation resources, we used the Babel Fish translation service
provided by Altavista.com for translating German topics into Italian, with English as a pivot language. Then the
translated Italian topics were used for retrieving Italian documents from the Italian document collection. The
translated Italian topics and the document collections were indexed using three different kinds of units: (1)
linguistically meaningful units, (2) character 6-grams, and (3) a combination of 1 and 2. We submitted three
automatic runs with the three indexing units.
1. Introduction
Clairvoyance participated in the CLEF 2003 bilingual retrieval track using the German and Italian language pair.
As we did not have German-to-Italian translation resources, we used the free Babel Fish translation service
provided by Altavista.com for translating German topics into Italian, with English as a pivot language. The
resulting translated Italian topics were used for retrieving Italian documents from the Italian document collection.
The translated Italian topics and the document collections were indexed using three different kinds of features:
(1) linguistically meaningful units (e.g., words and NPs), (2) character 6-grams, and (3) a combination of 1 and
2. We submitted three automatic runs, each based on one of the three indexing units. In the following sections,
we describe the details of our submission and present the performance results.
2. CLARIT Cross-Language Information Retrieval
In CLEF 2003, we adopted query translation as the means for bridging the language gap between the query
language and the document language for cross-language information retrieval. For German-to-Italian
information retrieval, first, a German query string was translated into Italian via machine translation; then the
translated Italian topics were used for retrieving Italian documents from the Italian document collection. For
query and document processing, we used the CLARIT system [1], in particular, those components encompassing
a newly developed Italian NLP module (for extracting Italian phrases), indexing (term weighting and phrasal
decomposition), retrieval, and “thesaurus extraction” (for extracting terms to support pseudo relevance
feedback).
2.1 Query Translation via a Pivot Language
The Babel Fish translation service (altavista.com) provides translation between selected language pairs including
German-to-English and English-to-Italian. It does not provide translation service between German and Italian
directly. So we used English as a pivot language, first translating the German topics to English and then
translating the English topics into Italian. As an illustration of typical results for this process, Figure 1 provides
the translations from Babel Fish for Topic 141.
Even though there was increased degradation in query quality after translation, we felt that, except for translation
of proper names, the quality of the translation from German to English and from English to Italian by Babel Fish
was adequate for the purpose of cross-language information retrieval. We quantitatively evaluate this impression
in Section 3.
2.2 Italian Topic Processing
Once the topics were translated into Italian, we extracted two types of terms from the topics: (1) linguistically
meaningful units or character n-grams.
To extract linguistically meaningful units, we used CLARIT Italian NLP. This NLP module makes use of a
lexicon and finite-state grammar for extracting phrases such as NPs. The lexicon is based on the Multext Italian
(1) Original German topic from CLEF-2003:
Briefbombe für Kiesbauer .
Finde Informationen über die Explosion einer Briefbombe im Studio der Moderatorin Arabella Kiesbauer beim
Fernsehsender PRO7.
(2) English translation of (1) by Babel Fish:
Letter bomb for gravel farmer. Find information about the explosion of a letter bomb in the studio of the host
Arabella gravel farmer with the television station PRO7.
(3) Ideal English topic from CLEF-2003:
Letter Bomb for Kiesbauer
Find information on the explosion of a letter bomb in the studio of the TV channel PRO7 presenter Arabella
Kiesbauer.
(4) Italian translation of (2) by Babel Fish:
Bomba della lettera per il coltivatore della ghiaia. Trovi le informazioni sull'esplosione di una bomba della lettera
nell'studio del coltivatore della ghiaia di Arabella ospite con la stazione PRO7 della televisione.
(5) Ideal Italian topic from CLEF-2003:
Lettera Bomba per Kiesbauer
Recupera le informazioni relative all'esplosione di una lettera bomba nello studio della presentatrice della rete
televisiva PRO7.
Figure 1: Topic 141 and its translations from Babel Fish
lexicon1, which was expanded by adding punctuations and special characters. In addition, entries with accented
vowels were duplicated by substituting the accented vowels with their corresponding unaccented vowels
followed by an apostrophe (“’”). The final lexicon contained about 135,000 entries. An Italian stop word list2,
which contained 433 entries, was used to filter out stop words. The grammar specified the rules for constructing
phrases, especially NPs, and morphological normalization rules for normalizing morphological variants to their
root forms, e.g., “previsto” to “prever”. In CLEF 2003 experiments, we extracted Adjectives, Verbs, and NPs as
indexing terms.
Another way to construct terms is to use overlapping character n-grams. We have observed that our lexicon-
based term extraction did not have complete coverage for morphological normalization. The n-gram approach
we adopted was aimed at mitigating such an effect. For the submissions, we have used overlapping 6-grams, as
it was previously reported to be effective [2]. Spaces and punctuations were included in the character 6-grams.
2.3 CLARIT Indexing and Retrieval
CLARIT indexing involves statistical analysis of a text corpus and construction of an inverted index, with each
index entry specifying the index word and a list of texts. CLARIT allows the index to be built upon full
documents or variable-length subdocuments. We used subdocuments as the basis for indexing and document
scoring in our experiments. The size of a subdocument was in the range of 8 sentences to 12 sentences.
CLARIT retrieval is based on the vector space retrieval model. Various similarity measures are supported in the
model. For CLEF 2003, we used the dot product function for computing similarities between a query and a document:
sim ( P , D ) = W P ( t ) ⋅ W D ( t ).
t∈ P ∩ D
where WP(t) is the weight associated with the query term t and WD(t) is the weight associated with the term t in
the document D. The two weights were computed as follows:
1
http://www.lpl.univ-aix.fr/projects/multext/LEX/LEX.SmpIt.html
2
Obtained from http://www.unine.ch/Info/clef/
W D ( t ) = TF D ( t ) ⋅ IDF ( t ).
W P ( t ) = C ( t ) ⋅ TF P ( t ) ⋅ IDF (t )
where IDF and TF are standard inverse document frequency and term frequency statistics, respectively. IDF(t)
was computed with the target corpus for retrieval. The coefficient C(t) is an “importance coefficient”, which can
be modified either manually by the user or automatically by the system (e.g., updated during feedback).
2.4 Post-Translation Query Expansion
Query expansion through (pseudo) relevance feedback has proved to be effective for improving IR performance
[3]. We used pseudo relevance feedback for augmenting the queries. After retrieving some documents for a
given topic from the target corpus, we took a set of top ranked documents, regarding them as relevant documents
to the query, and extracted terms from the these documents. The terms were ranked based on the following
formula:
N −R + 2 R +1
Prob2(t) = log(R + 1) x log( − 1) − log( − 1)
Nt − Rt +1
t
Rt
where N is the number of documents in the target corpus, Nt is the number of documents in the corpus that
contain term t, R is the number of documents for feedback that are (presumed to be) relevant to the topic, and Rt
is the number of documents that are (presumed to be) relevant to the topic and contain term t.
3 Experiments
We submitted three automatic runs to CLEF 2003. All the queries used the title and description fields
(Ttitle+Description) of the topics provided by CLEF 2003. The results presented below are based on relevance
judgments of 42 topics, which have relevant documents in the Italian corpus. The three runs were:
• ccwrd: with linguistically meaningful units as indexing terms
• ccngm: with character 6-grams as indexing terms
• ccmix: a combination of linguistic units and character 6-grams as indexing units
With the ccmix run, the combinations were constructed through a simple concatenation of the terms nominated
by ccwrd and ccngm. We ran Italian monolingual experiments to obtain the baseline with ideal translations
after obtaining the relevance judgments from CLEF 2003.
All the experiments were run with post-translation pseudo relevance feedback. The feedback-related parameters
were based on training over CLEF 2002 topics. The settings for German-to-Italian retrieval were: extracting
T=80 terms from the top N=25 retrieved documents with the Prob2 method. For the n-gram based indexing and
the mixed model, an additional term cutoff percentage set to P=0.01. For the word-based indexing, the
percentage cutoff is set to P=0.25. For Italian monolingual retrieval with words as indexing terms: T=50, N=50,
P=0.1. For Italian monolingual retrieval with n-grams and the mixed model as indexing terms: T=80, N=25,
P=0.05.
Table 1 presents the results for our submitted runs, and Table 2 presents results for our training runs with CLEF
2002 topics. The monolingual baselines for the ccwrd*, ccngm*, and ccmix* runs are the optimal monolingual
runs based on word based indexing, 6-gram based indexing, and mixed indexing, respectively. While the 6-gram
based indexing produced higher average precision compared with the word based indexing for the CLEF 2002
topics, it significantly underperformed word based indexing for CLEF 2003 topics. Further examination is
required to account for the difference in behavior of the two indexing methods for the two top sets.
Run ID Indexing Units Recall AP % mono AP
ccwrd Adj+VP+NPs 541/809 0.2303 67.2% (of 0.3428)
ccngm Character 6-grams 456/809 0.1624 54.3% (of 0.2993)
ccmix Adj+VP+NPs, character 6-grams 505/809 0.2098 61.7% (of 0.3402)
Table 1: German-to-Italian retrieval performance with CLEF 2003 topics. All three runs are our submitted runs.
Run ID Indexing Units Recall AP % mono AP
ccwrd2002 Adj+VP+NPs 643/1072 0.1823 61.6% (of 0.2959)
ccngm2002 Character 6-grams 648/1072 0.2133 68.1% (of 0.3132)
ccmix2002 Adj+VP+NPs, character 6-grams 675/1072 0.2147 60.1% (of 0.3574)
Table 2: German-to-Italian retrieval performance with CLEF 2002 topics.
Topics 2002 Avg. Prec 2003 Avg. Prec
(1) Translated English (from German) to Italian 0.1549 0.1748
Performance change compared with (2) (-24.2%) (-16.1%)
Performance change compared with (3) (-36.7%) (-45.3%)
(2) Ideal English to Italian 0.2048 0.2083
Performance change compared with (3) (-16.4%) (-34.8%)
(3) Ideal Italian 0.2449 0.3197
Table 3: Performance comparison between different versions of topics
Table 3 presents a comparison between different versions of the Italian topics for CLEF 2002 and CLEF 2003
topics. The average precision statistics were computed with word based indexing and with no feedback. Even
through comparing different topic statements is not justified methodologically [4], the comparison gives us a
rough estimate of the quality of the translation module. Translation from English to Italian decreased
performance in the range of 16.4% to 34.8%, while adding another layer of translation from German to English
decreased performance further by 16.1% to 24.2%. This shows that translation service such as Babel Fish still
needs to be improved for better CLIR performance.
4 Conclusions
Due to the lack of resources, our participation in CLEF 2003 was limited. We succeeded in submitting three
runs for German-to-Italian retrieval, examining word based indexing and n-gram based indexing. Our results
with CLEF 2002 and CLEF 2003 did not provide firm evidence of which indexing method is better. Future
analysis is required in this direction.
References
[1] Evans, D.A., and R.G. Lefferts. CLARIT–TREC Experiments. Information Processing and Management,
Vol.31, No.3, pp.385–395, 1995.
[2] McNamee, P., and J. Mayfield. Scalable Multilingual Information Access. In C. Peters, editor, Working
Notes for the CLEF 2002 Workshop, pp.133–140, 2002.
[3] Ballesteros, L., and W. B. Croft. Statistical Methods for Cross-Language Information Retrieval. In G.
Grefenstette, editor, Cross-Language Information Retrieval, Chapter 3. Kluwer Academic Publishers, Boston,
pp.23–40, 1998.
[4] Voorhees, E. The Philosophy of Information Retrieval Evaluation. In C. Peters, M. Braschler, J. Gonzalo,
and M. Kluck, editors, Evaluation of Cross-Language Information Retrieval Systems: Proceedings of the CLEF
2001 Workshop, Lecture Notes in Computer Science 2406, Springer, pp.355–370, 2001.