BoostCLIR: JP-EN Relevance Marked Patent Corpus (ICPSR doi:10.11588/data/10001)

View:

Part 1: Document Description
Part 2: Study Description
Part 5: Other Study-Related Materials
Entire Codebook

Document Description

Citation

Title:

BoostCLIR: JP-EN Relevance Marked Patent Corpus

Identification Number:

doi:10.11588/data/10001

Distributor:

heiDATA

Date of Distribution:

2014-06-16

Version:

1

Bibliographic Citation:

Sokolov, Artem; Jehl Laura; Hieber Felix; Ruppert, Eugen; Riezler, Stefan, 2014, "BoostCLIR: JP-EN Relevance Marked Patent Corpus", https://doi.org/10.11588/data/10001, heiDATA, V1

Study Description

Citation

Title:

BoostCLIR: JP-EN Relevance Marked Patent Corpus

Identification Number:

doi:10.11588/data/10001

Authoring Entity:

Sokolov, Artem (Department of Computational Linguistics)

Jehl Laura (Department of Computational Linguistics)

Hieber Felix (Department of Computational Linguistics)

Ruppert, Eugen (Department of Computational Linguistics)

Riezler, Stefan (Department of Computational Linguistics)

Producer:

Jehl, Laura

Sokolov, Artem

Ruppert, Eugen

Date of Production:

2013

Distributor:

heiDATA

Date of Distribution:

2014-06-16

Study Scope

Keywords:

Computer and Information Science

Abstract:

BoostCLIR is a bilingual (Japanese-English) corpus of patent abstracts, extracted from the <a href='http://www.ifs.tuwien.ac.at/imp/marec.shtml'>MAREC</a> patent data, and the data from the <a href='http://research.nii.ac.jp/ntcir/data/data-en.html'>NTCIR PatentMT workshop</a> collections, accompanied with relevance judgements for the task of patent prior-art search. <br /><br /> <strong>Important:</strong> The English side of the corpus contains patent IDs as well as the text of the abstracts. The Japanese side only contains patent IDs because of NTCIR copyright restrictions. The Jap anese patent abstracts can be extracted from full text Japanese patent documents, which are available from the organizers of the NTCIR workshop. <br /><br /> The corpus contains training, development and testing subsets sampled from non-intersecting time periods. <br /><br /> Relevance judgement for patent retrieval are constructed from patent citations by assigning three integer levels to three categories of relationships, with highest relevance (3) for family patents, lower relevance for patents cited in search reports by patent examiners (2), and lowest relevance level (1) for applicants’ citations. <br /><br /> For a detailed descrip tion of the corpus construction process, please see the above publication.

Kind of Data:

textual data

Methodology and Processing

Other Study-Related Materials

Label:

boostclir.tar.gz

Text:

data set

Notes:

application/x-gzip

Other Study-Related Materials

Label:

README_BoostCLIR.txt

Text:

README

Notes:

text/plain; charset=US-ASCII