LDC2010T07* **- **Chinese Treebank 7.0 <#ctb> -*
LDC2010T11* **- * *NIST 2003 Open Machine Translation (OpenMT) Evaluation* <#open>* -*
LDC2010V01* **- TRECVID 2004 Keyframes & Transcripts <#trecvid>** -*
------------------------------------------------------------------------
*Mark Liberman, LDC Director, wins the 2010 Antonio Zampolli Prize*
LDC is proud to announce that our founder and Director, Mark Liberman, was awarded the 2010 Antonio Zampolli prize <http://www.elra.info/Antonio-Zampolli-Prize.html> at LREC2010 <http://www.lrec-conf.org/lrec2010/>, hosted by ELRA <http://www.elra.info/>, the European Language Resource Association. This prestigious honor is given by ELRA's board members to recognize "outstanding contributions to the advancement of language resources and language technology evaluation within human language technologies".
Mark's prize talk, delivered on May 21, 2010 and entitled The Future of Computational Linguistics: or, What Would Antonio Zampolli Do? <http://languagelog.ldc.upenn.edu/myl/AntonioZampolliPrizeLecture.pdf>, discussed Antonio Zampolli's far-reaching contributions to the language technology community and how his vision resonates in Mark's research. Please join us in congratulating Mark on receiving this award.
[ top <#top>]
*New Publications*
(1) C <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2010T07>hinese Treebank 7.0 <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2010T07> consists of 840,000 words of annotated and parsed text from Chinese newswire, magazine news, and various broadcast news and broadcast conversation programs. The Chinese Treebank project began at the University of Pennsylvania in 1998, continued at the University of Colorado, and is in the process of moving to Brandeis University <http://www.cs.brandeis.edu/%7Ellc/page2/page2.html>. The project provides a large, part-of-speech tagged and fully bracketed Chinese language corpus. The first deliveries provided syntactically annotated words from newswire texts. The annotation of broadcast news and broadcast conversation data began and continues under the DARPA GALE (Global Autonomous Language Exploitation) program; Chinese Treebank 7.0 represents the results of that effort.
Chinese Treebank 7.0 includes text from the following genres and sources.
*Genre*
*# words*
Newswire (Xinhua)
250,000
News Magazine (Sinorama)
150,000
Broadcast News (CBS, CNR, CTS, CCTV, VOM)
270,000
Broadcast Conversation (CCTV, CNN, MSNBC, Phoenix)
170,000
Total
840,000
The annotation of syntactic structure trees for the Chinese newswire data was taken from Chinese Treebank 5.0 and updated with some corrections. Known problems, like multiple tree nodes at the top level, were fixed. Inconsistent annotations for object control verbs were also corrected. The residual Traditional Chinese characters in the Sinorama portion of the data, the result of incomplete automatic conversion, have been manually normalized to Simplified Chinese characters.
This release contains the frame files for each annotated verb or noun, which specify the argument structure (semantic roles) for each predicate. The frame files are effectively lexical guidelines for the propbank annotation. The semantic roles annotated in this data can only be interpreted with respect to these frame files. The annotation of the verbs in the Xinhua news portion of the data is taken from Chinese Proposition Bank 1.0 (LDC2005T23) <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2005T23>. The annotation of the predicate-argument structure of the included nouns, which are primarily nominalizations, has not been previously released. The Sinorama portion of the data, both for verbs and nouns, has not been previously released.
[ top <#top>]
*
(2) NIST 2003 Open Machine Translation (OpenMT) Evaluation <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2010T11> is a package containing source data, reference translations, and scoring software used in the NIST 2003 OpenMT evaluation. It is designed to help evaluate the effectiveness of machine translation systems. The package was compiled and scoring software was developed by researchers at NIST, making use of newswire source data and reference translations collected and developed by LDC.
The objective of the NIST OpenMT evaluation series is to support research in, and help advance the state of the art of, machine translation (MT) technologies -- technologies that translate text between human languages. Input may include all forms of text. The goal is for the output to be an adequate and fluent translation of the original. Additional information about these evaluations may be found at the NIST Open Machine Translation (OpenMT) Evaluation web site <http://www.itl.nist.gov/iad/mig/tests/mt/>.
This evaluation kit includes a single perl script that may be used to produce a translation quality score for one (or more) MT systems. The script works by comparing the system output translation with a set of (expert) reference translations of the same source text. Comparison is based on finding sequences of words in the reference translations that match word sequences in the system output translation.
The Chinese-language and Arabic-language source text included in this corpus is a reorganization of data that was initially released to the public respectively as Multiple-Translation Chinese (MTC) Part 4 (LDC2006T04) <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2006T04> and Multiple-Translation Arabic (MTA) Part 2 (LDC2005T05) <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2005T05>. The reference translations are a reorganized subset of data from these same Multiple-Translation corpora. All source data for this corpus is newswire text collected in January and February of 2003 from Agence France-Presse, and Xinhua News Agency. For details on the methodology of the source data collection and production of reference translations, see the documentation for the above-mentioned corpora.
For each language, the test set consists of two files, a source and a reference file. Each reference file contains four independent translations of the data set. The evaluation year, source language, test set, version of the data, and source vs. reference file are reflected in the file name.
[ top <#top>]
*
(3) TRECVID 2004 Keyframes and Transcripts <http://www.ldc.upenn.edu/Catalog/CatalogEntry.jsp?catalogId=LDC2010V01> was developed as a collaborative effort between researchers at LDC, NIST <http://www.nist.gov/>, LIMSI-CNRS <http://www.limsi.fr/>, and Dublin City University <http://www.dcu.ie/>. TREC Video Retrieval Evaluation (TRECVID) is sponsored by the National Institute of Standards and Technology (NIST) to promote progress in content-based retrieval from digital video via open, metrics-based evaluation. The keyframes in this release were extracted for use in the NIST TRECVID 2004 Evaluation. TRECVID is a laboratory-style evaluation that attempts to model real world situations or significant component tasks involved in such situations. In 2004 there were four main tasks with associated tests:
* shot boundary determination
* story segmentation
* high-level feature extraction
* search (interactive and manual)
For a detailed description of the TRECVID Evaluation Tasks, please refer to the NIST TRECVID 2004 Evaluation Description. <http://www-nlpir.nist.gov/projects/tv2004/>
The source data includes approximately 70 hours of English language broadcast programming collected by LDC in 1998 from ABC ("World News Tonight") and CNN ("CNN Headline News").
Shots are fundamental units of video, useful for higher-level processing. To create the master list of shots, the video was segmented. The results of this pass are called subshots. Because the master shot reference is designed for use in manual assessment, a second pass over the segmentation was made to create the master shots of at least 2 seconds in length. These master shots are the ones used in submitting results for the feature and search tasks in the evaluation. In the second pass, starting at the beginning of each file, the subshots were aggregated, if necessary, until the current shot was at least 2 seconds in duration, at which point the aggregation began anew with the next subshot.
The keyframes were selected by going to the middle frame of the shot boundary, then parsing left and right of that frame to locate the nearest I-Frame. This then became the keyframe and was extracted. Keyframes have been provided at both the subshot (NRKF) and master shot (RKF) levels.
[ top <#top>]
------------------------------------------------------------------------
Ilya Ahtaridis Membership Coordinator -------------------------------------------------------------------- Linguistic Data Consortium Phone: (215) 573-1275 University of Pennsylvania Fax: (215) 573-2175 3600 Market St., Suite 810 ldc at ldc.upenn.edu Philadelphia, PA 19104 USA http://www.ldc.upenn.edu
-------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: text/html Size: 13796 bytes Desc: not available URL: <http://www.uib.no/mailman/public/corpora/attachments/20100624/646444dc/attachment.txt>