Pseudo Relevance Feedback Method Based on Taylor Expansion of Re-
trieval Function in NTCIR-3 Patent Retrieval Task 
 
Kazuaki KISHIDA 
Faculty of Cultural Information Resources 
Surugadai University 
698 Azu, Hanno, Saitama 357-8555 JAPAN 
kishida@surugadai.ac.jp 
 
 
Abstract 
Pseudo relevance feedback is empirically 
known as a useful method for enhancing 
retrieval performance. For example, we 
can apply the Rocchio method, which is 
well-known relevance feedback method, 
to the results of an initial search by as-
suming that the top-ranked documents are 
relevant. In this paper, for searching the 
NTCIR-3 patent test collection through 
pseudo feedback, we employ two rele-
vance feedback mechanism; (1) the Roc-
chio method, and (2) a new method that is 
based on Taylor formula of linear search 
functions. The test collection consists of 
near 700,000 records including full text of 
Japanese patent materials. Unfortunately, 
effectiveness of our pseudo feedback 
methods was not empirically observed at 
all in the experiment.  
1 Introduction 
Relevance feedback is widely recognized as an 
effective method for improving retrieval effective-
ness in the context of interactive IR. As often 
pointed out, it is difficult for users to represent 
their own information needs into a well-defined set 
of search terms or statements. The resulting short 
or poor queries would bring them unsatisfactory 
results. However, if a few relevant documents hap-
pen to be found by the search, we could automati-
cally or manually extract some useful terms from 
the documents, and add them to the initial search 
expression. It is obviously expected that search 
effectiveness of the second search using the ex-
tended query will be improved significantly. This 
is a basic idea of relevance feedback. 
Inevitably, for executing automatic relevance 
feedback, the system has to obtain relevance in-
formation, i.e., relevant or irrelevant documents, 
from the users interactively. However, some re-
searchers have tried to employ relevance feedback 
techniques with no relevance information. The ob-
jective is to enhance search performance of re-
trieval models such as vector space model, 
probabilistic model and so on, without interaction 
on relevance information between system and us-
ers. The technique is usually called pseudo rele-
vance feedback, in which a standard feedback 
method (e.g., the Rocchio method) is applied by 
assuming that top-ranked documents searched by 
the initial search are relevant. 
The purpose of this paper is to report results of 
retrieval experiments for examining effectiveness 
of pseudo relevance feedback in the case of search-
ing a patent collection. In particular, we attempt to 
compare search performance of the traditional 
Rocchio method with that of an alternative method, 
which is based on Taylor approximation of re-
trieval function proposed by Kishida[1]. This re-
port is based on two experiments using the 
NTCIR-1 test collection and the NTCIR-3 patent 
test collection, respectively. As to the latter, the 
results were obtained at the time of NTCIR-3 
Workshop held in October 2002 [2]. 
The rest of this paper is organized as follows. In 
Section 2, the Rocchio method and an alternative 
method proposed by Kishida[1] will be introduced. 
In Section 3 a preliminary experiment for confirm-
ing how well the alternative method works in a 
normal relevance feedback situation will be de-
scribed. The NTCIR-1 test collection with rele-
vance judgment information is used for the 
preliminary experiment. In Section 4, results of an 
experiment on pseudo relevance feedback method 

References

[1] K. Kishida. 2001. Feedback method for docu-ment retrieval using numerical values on rele-vance given by users. IPSJ SIG Notes Fundamental Infology, 61: 189-196. (in Japa-nese)

[2] K. Kishida. 2003. Experiment on Pseudo Rele-vance Feedback Method Using Taylor Formula at NTCIR-3 Patent Retrieval Task. Proceed-ings of the Third NTCIR Workshop on Re-search in Information Retrieval, Automatic Text Summarization and Question Answering, NII, Tokyo. http://research.nii.ac.jp/ntcir/

[3] J. J. Rocchio, Jr. 1971. Relevance feedback in information retrieval. in G. Salton ed., The SMART Retrieval System: Experiments in Automatic Document Processing, Prentice-Hall, Englewood Cliffs, NJ, 313-323.

[4] G. Salton and C. Buckley. 1990. Improving retrieval performance by relevance feedback. Journal of the American Society for Informa-tion Science, 41: 288-297.

[5] P. Sarinivasan. 1996. Query expansion and MEDLINE. Information Processing and Management, 32: 431-443.

[6] J. H. Lee. 1998. Combining the evidence of different relevance feedback methods for in-formation retrieval. Information Processing and Management, 34: 681-691

[7] R. Mandala, T. Tokunaga and H. Tanaka. 2000. Query expansion using heterogeneous thesauri. Information Processing and Management, 36: 361-378.

[8] M. Iwayama. 2000. Relevance feedback with a small number of relevance judgments: incre-mental relevance feedback vs. Document clus-tering. in Proceedings of the 23rd Annual International ACM SIGIR Conference on Re-search and Development in Information Re-trieval, ACM Press, 10-16

[9] G. Ciocca. and R. Schettini.1999. A relevance feedback mechanism for content-based image retrieval. Information Processing and Man-agement, 35: 605-632.

[10] M. F. Moens and J. Dumortier. 2000. Text categorization: the assignment of subject de-scriptors to magazine articles. Information Processing and Management, 36: 841-861.

[11] C. Buckley, J. Allan, and G. Salton. 1994. Automatic routing and ad-hoc retrieval using SMART: TREC2. in D.K. Harman ed., The Second Text Retrieval Conference (TREC2). National Institute of Standards and Technology, Gaithersburg MD, 45-55.

[12] S. E. Robertson, et al. 1995. Okapi at TERC-3. in D.K. Harman ed. Overview of the Third Text Retrieval Conference (TREC-3). National Institute of Standards and Technology, Gaithersburg MD, 109-126.

[13] D. A. Harville. 1997. Matrix Algebra from a Statisticians Perspective. Springer, New York.

[14] Yuji Matsumoto, Akira Kitauchi, Tatsuo Yamashita, Yoshitaka Hirano, Hiroshi Matsuda, Kazuma Takaoka and Masayuki Asahara. 2000. Morphological Analysis System ChaSen version 2.2.1 Manual. http://chasen.aist-nara.ac.jp/
