Friday, June 17, 2011

LDC June 2011 Newsletter




ACL has returned to North America and LDC is taking this opportunity to interact with top HLT researchers in beautiful Portland, OR. LDC’s exhibition table will feature information on new developments at the consortium and will also be the go-to point for exciting new, green giveaways.

LDC’s Seth Kulick will be presenting research on ‘Using Derivation Trees for Treebank Error Detection’ (S-66) during Monday’s evening poster session (20 June, 6.00 – 8.30 pm). The abstract for this paper, coauthored by LDCers Ann Bies and Justin Mott, is as follows:

This work introduces a new approach to checking treebank consistency. Derivation trees based on a variant of Tree Adjoining Grammar are used to compare the annotation of word sequences based on their structural similarity. This overcomes the problems of earlier approaches based on using strings of words rather than tree structure to identify the appropriate contexts for comparison. We report on the result of applying this approach to the Penn Arabic Treebank and how this approach leads to high precision of error detection.

We hope to see you there.

LDC is now on your favorite Social Networks (Facebook, LinkedIn and RSS, oh my!)

Over the past few months, LDC has responded to requests from the community to increase our online presence. We are happy to announce that LDC now has its very own Facebook page, LinkedIn profile (independent of the University of Pennsylvania) and Blog, which provides an RSS feed for LDC newsletters. Please visit LDC on our various profiles and let us know what you think!


New Publications

(1) 2006 NIST Spoken Term Detection Development Set was compiled by researchers at NIST (National Institute of Standards and Technology) and contains eighteen hours of Arabic, Chinese and English broadcast news, English conversational telephone speech and English meeting room speech used in NIST's 2006 Spoken Term Detection (STD) evaluation. The STD initiative is designed to facilitate research and development of technology for retrieving information from archives of speech data with the goals of exploring promising new ideas in spoken term detection, developing advanced technology incorporating these ideas, measuring the performance of this technology and establishing a community for the exchange of research results and technical insights.
The 2006 STD task was to find all of the occurrences of a specified term (a sequence of one or more words) in a given corpus of speech data. The evaluation was intended to develop technology for rapidly searching very large quantities of audio data. Although the evaluation used modest amounts of data, it was structured to simulate the very large data situation and to make it possible to extrapolate the speed measurements to much larger data sets. Therefore, systems were implemented in two phases: indexing and searching. In the indexing phase, the system processes the speech data without knowledge of the terms. In the searching phase, the system uses the terms, the index, and optionally the audio to detect term occurrences.

The development corpus consists of three data genres: broadcast news (BN), conversational telephone speech (CTS) and conference room meetings (CONFMTG). The broadcast news material was collected in 2001 by LDC's broadcast collection system from the following sources: ABC (English), China Broadcasting System (Chinese), China Central TV (Chinese), China National Radio (Chinese), China Television System (Chinese), CNN (English), MSNBC/NBC (English), Nile TV (Arabic), Public Radio International (English) and Voice of America (Arabic, Chinese, English). The CTS data was taken from the Switchboard data sets (e.g., Switchboard-2 Phase 1 LDC98S75, Switchboard-2 Phase 2 LDC99S79) and the Fisher corpora (e.g., Fisher English Training Sppech Part 1 LDC2004S13), also collected by LDC. The conference room meeting material consists of goal-oriented, small group round table meetings and was collected in 2001, 2004 and 2005 by NIST, the International Computer Science Institute (Berkeley, California), Carnegie Mellon University (Pittsburgh, PA) and Virginia Polytechnic Institute and State University (Blacksburg, VA) as part of the AMI corpus project.

Each BNews recording is a 1-channel, pcm-encoded, 16Khz, SPHERE formatted file. CTS recordings are 2-channel, u-law encoded, 8 Khz, SPHERE formatted files. TheCONFMTG files contain a single recorded channel.

2006 NIST Spoken Term Detection Development Set is distributed on 1 DVD-ROM. 2011 Subscription Members will automatically receive two copies of this corpus. 2011 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for US$800.

*

(2) Datasets for Generic Relation Extraction (reACE) was developed at The University of Edinburgh, Edinburgh, Scotland. It consists of English broadcast news and newswire data originally annotated for the ACE (Automatic Content Extraction) program to which the Edinburgh Regularized ACE (reACE) mark-up has been applied.

The Edinburgh relation extraction (RE) task aims to identify useful information in text (e.g., PersonW works for OrganisationX, GeneY encodes ProteinZ) and to recode it in a format such as a relational database or RDF triple store (a database for the storage and retrieval of Resource Description Framework (RDF) metadata) that can be more effectively used for querying and automated reasoning. A number of resources have been developed for training and evaluation of automatic systems for RE in different domains. However, comparative evaluation is impeded by the fact that these corpora use different markup formats and different notions of what constitutes a relation.

reACE solves this problem by converting data to a common document type using token standoff and including detailed linguistic markup while maintaining all information in the original annotation. The subsequent re-annotation process normalizes the two data sets so that they comply with a notion of relation that is intuitive, simple and informed by the semantic web.

The data in this corpus consists of newswire and broadcast news material from ACE 2004 Multilingual Training Corpus LDC 2005T09 and ACE 2005 Multilingual Training Corpus LDC2006T06 . This material has been standardized for evaluation of multi-type RE across domains.

Annotation includes (1) a refactored version of the original data to a common XML document type; (2) linguistic information from LT-TTT (a system for tokenizing text and adding markup) and MINIPAR (an English parser); and (3) a normalized version of the original RE markup that complies with a shared notion of what constitutes a relation across domains.

The data sources represented in the corpus were collected by LDC in 2000 and 2003 and consist of the following: ABC, Agence France Presse, Associated Press, Cable News Network, MSNBC/NBC, New York Times, Public Radio International, Voice of America and Xinhua News Agency.

Datasets for Generic Relation Extraction (reACE) is distributed via web download. 2011 Subscription Members will automatically receive two copies of this corpus on disc. 2011 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for US$800.

*

(3) English Gigaword Fifth Edition is a comprehensive archive of newswire text data that has been acquired over several years by the LDC at the University of Pennsylvania. The fifth edition includes all of the contents in English Gigaword Fourth Edition (LDC2009T13) plus new data covering the 24-month period of January 2009 through December 2010.

The seven distinct international sources of English newswire included in this edition are the following:
  • Agence France-Presse, English Service (afp_eng)
  • Associated Press Worldstream, English Service (apw_eng)
  • Central News Agency of Taiwan, English Service (cna_eng)
  • Los Angeles Times/Washington Post Newswire Service (ltw_eng)
  • Washington Post/Bloomberg Newswire Service (wpb_eng)
  • New York Times Newswire Service (nyt_eng)
  • Xinhua News Agency, English Service (xin_eng)
The seven letter codes in the parentheses above include the three-character source name abbreviations and the three-character language code ("eng") separated by an underscore ("_") character. The three-letter language code conforms to LDC's internal convention based on the ISO 639-3 standard.

Data

The following table sets forth the overall totals for each source. Note that "Total-MB" refers to the quantity of date when unzipped (approximately 26 gigabytes), "Gzip-MB" refers to compressed file sizes as stored on the DVD-ROMs and "K-wrds" refers to the number of whitespace-separated tokens (of all types) after all SGML tags are eliminated:

Source
#Files
Gzip-MB
Totl-MB
K-wrds
#DOCs




afp_eng
146
1732
4937
738322
2479624

apw_eng
193
2700
7889
1186955
3107777

cna_eng
144
86
261
38491
145317

ltw_eng
127
651
1694
268088
411032

nyt_eng
197
3280
8938
1422670
1962178

wpb_eng
12
42
111
17462
26143

xin_eng
191
834
2518
360714
1744025




TOTAL
1010
9325
26348
4032686
9876086

English Gigaword Fifth Edition is distributed on 3 DVD-ROM. 2011 Subscription Members will automatically receive two copies of this corpus. 2011 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for US$6000.