- 

In this newsletter:

- Fall 2010 LDC Data Scholarship Winners! -

- Position Openings at LDC -

New Publications:

LDC2010T18

- ACE Time Normalization (TERN) 2004 English Evaluation Data V1.0 -

LDC2010T19
- Korean Newswire Second Edition -

LDC2010T17
- NIST 2006 Open Machine Translation (OpenMT) Evaluation -

Fall 2010 LDC Data Scholarship Winners!

LDC is pleased to announce the winners in our first-ever LDC Data Scholarship program!  The LDC Data Scholarship program provides university students with access to LDC data at no-cost.  Data scholarships are offered twice a year to correspond to the Fall and Spring semesters.  Students are asked to complete an application which consists of a data use proposal and letter of support from their academic adviser.  

LDC received many strong applications from both undergraduate and graduate students attending universities across the globe.  The decision process was difficult, and after much deliberation, we have selected eight winners!   These students will receive no-cost copies of LDC data valued at over US$10,000:

Aby Abraham - Ohio University (USA), graduate student, Electrical Engineering.  Aby has been awarded a copy of 2003 NIST Speaker Recognition Evaluation (LDC2010S03) for his work in using long term memory cells for continuous speech recognition.

Ripandy Adha - Bandung Institute of Technology (Indonesia), undergraduate student, Computer Science - Ripandy has been awarded a copy of American English Spoken Lexicon (LDC99L23) to assist in the development of a voice command internet browser.

Basawaraj - Ohio University (USA), PhD candidate, Electrical Engineering and Computer Science.  Basawaraj has been awarded a copy of NIST 2002 Open Machine Translation (OpenMT) Evaluation (LDC2010T10) to assist in fine tuning his machine translation system and to provide a benchmark dataset.

Zachary Brooks - University of Arizona (USA), PhD Candidate, Second Language Acquisition and Teaching.  Zachary and his research group have been awarded a copy of ECI Multilingual Text (LDC94T5) for research in eye movement tracking by native and non-natives readers.

Marco Carmosino - Hampshire College (USA), undergraduate student, Computer Science.  Marco has been awarded a copy of English Gigaword Fourth Edition (LDC2009T13) for his work in narrative chain extraction.

Xiaohui Huang - Harbin Institute of Technology (China), Shenzhen Graduate School.  Xiaohui has been awarded a copy of TDT5 Topics and Annotations (LDC2006T19)  for his work in topic detection and tracking for large-scale web  data.

Yuhuan Zhou - PLA University of Science and Technology (China), postgraduate student, Institute of Communications Engineering.  Yuhuan has been awarded a copy of 2002 NIST Speaker Recognition Evaluation (LDC2004S04) to assist in the development of a speaker recognition system which fuses support vector data description (SVDD) and Gaussian mixture model (GMM).

Speaker Recognition Group (GEDA) with members Matias Fineschi, Gonzalo Lavigna, Jorge Prendes, and Pablo Vacatello -  Buenos Aires Institute of Technology (Argentina), Department of Electrical Engineering.  GEDA has been awarded a copy of 2004 NIST Speaker Recognition Evaluation (LDC2006S44) to assist in the development of a flexible platform on speaker verification capable of implementing different feature extraction, normalizations, stochastical models and outputs.

Please join us in congratulating our student winners!   The next LDC Data Scholarship program is scheduled for the Spring 2011 semester. Stay tuned for further announcements.

[ top ]

 Position Openings at LDC

Linguistic Data Consortium at the University of Pennsylvania has a number of immediate openings for full-time positions to support our corpus development projects:

• PROGRAMMER ANALYST - (#100528459 and #100929195)

Support linguistic data collection and annotation projects by providing software development, system integration, technical and research support, annotation tool development and/or data collection system management.

• SENIOR PROJECT MANAGER (#100728923 and #100728924)

Provide complete oversight for multiple, concurrent corpus creation projects, including collection, annotation and distribution of speech, text and/or video data in a variety of languages. Create project roadmaps and direct teams of programmers, linguists and managers to execute deliverables; represent corpus creation efforts to external researchers and sponsors.

• LEAD ANNOTATOR (#100728920)

Perform linguistic annotation on English text, speech and video data; recruit, train and supervise teams of annotators for multiple tasks and languages; define, test and document procedural approaches to linguistic annotation;perform quality control on annotated data.

For further information on the duties and qualifications for these positions, or to apply online please visit https://jobs.hr.upenn.edu/; search postings for the reference numbers indicated above.

Penn offers an excellent benefits package including medical/dental, retirement plans, tuition assistance and a minimum of three weeks paid vacation per year. The University of Pennsylvania is an affirmative action/equal opportunity employer.  All positions contingent upon grant funding.
.
For more information about LDC and the programs we support, visit http://www.ldc.upenn.edu/.

[ top ]

New Publications

(1) ACE Time Normalization (TERN) 2004 English Evaluation Data V1.0 was developed by researchers at The MITRE Corporation. It contains the English evaluation data prepared for the 2004 Time Expression Recognition and Normalization (TERN) Evaluation, sponsored by the Automatic Content Extraction (ACE) program, specifically, English broadcast news and newswire data collected by LDC. The training data for this evaluation can be found in ACE Time Normalization (TERN) 2004 English Training Data v 1.0 LDC2005T07.

The purpose of the TERN evaluation is to advance the state of the art in the automatic recognition and normalization of natural language temporal expressions. In most language contexts such expressions are indexical. For example, with 'Monday,' 'last week,' or 'three months starting October 1,' one must know the narrative reference time in order to pinpoint the time interval being conveyed by the expression. In addition, for data exchange purposes, it is essential that the identified interval be rendered according to an established standard, i.e., normalized. Accurate identification and normalization of temporal expressions are in turn essential for the temporal reasoning being demanded by advanced NLP applications such as question answering, information extraction and summarization.

The data in this release is English broadcast transcripts and newswire material from TDT4 Multilingual Text and Annotations LDC2005T16. The annotation specifications for this corpus were developed under DARPA's Translingual Information Detection Extraction and Summarization (TIDES) program, with support from ACE. All files have been doubly-annotated by two separate annotators and then reconciled, using the TIDES 2003 Standard for the Annotation of Temporal Expressions.  The data directory contains the corpus which consists of 192 files (54K words).

[ top ]

*

(2) Korean Newswire Second Edition is an archive of Korean newswire text that has been acquired over several years (1994-2009) at LDC from the Korean Press Agency. This release includes all of the content of Korean Newswire (LDC2000T45) (June 1994-March 2000) as well as newly-collected data.  The second edition contains all data collected by LDC from April 2000 through December 2009.

All material, including that from the first release, has been converted to UTF-8 (except for more recent data already in UTF-8 format) and processed in LDC's gigaword format. The gigaword format classifies newswire content into three types: story, multi and other where 'story' refers to an article containing information pertaining to a particular event on a day; 'multi' refers to an article that contains more than one story relating to different topics; and 'other' refers to articles containing lists, tables or numerical data, such as sports scores.

A word break error in the original release and in data collected from January 2002 through February 2005 has been corrected in the second edition with the result that all Korean text should display correctly. The error involved a line break in the middle of a word with the result that an affected word appeared in segments in two lines. This problem was  resolved using word histograms and a few

[ top ]

*

(3) NIST 2006 Open Machine Translation (OpenMT) Evaluation is a package containing source data, reference translations and scoring software used in the NIST 2006 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 broadcast, newswire and web newsgroup source data and reference translations collected and developed by LDC.

The objective of the NIST Open Machine Translation (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.  The OpenMT evaluations are intended to be of interest to all researchers working on the general problem of automatic translation between human languages. To this end, they are designed to be simple, to focus on core technology issues and to be fully supported. The 2006 task was to evaluate translation from Arabic to English and from Chinese to English.  Additional information about these evaluations may be found at the NIST Open Machine Translation (OpenMT) Evaluation web site.

This evaluation kit includes a single Perl script (mteval-v11b.pl) 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 included scoring script was released with the original evaluation, intended for use with SGML-formatted data files, and is provided to ensure compatibility of user scoring results with results from the original evaluation. An updated scoring software package (mteval-v13a-20091001.tar.gz), with XML support, additional options and bug fixes, documentation, and example translations, may be downloaded from the NIST Multimodal Information Group Tools website.

This release contains of 357 documents with corresponding sets of four separate human expert reference translations. The source data is comprised of Arabic and Chinese newswire documents, human transcriptions of broadcast news and broadcast conversation programs and web newsgroup documents collected by LDC in 2006. The newswire and broadcast material are from Agence France-Presse (Arabic, Chinese), Xinhua News Agency (Arabic, Chinese), Lebanese Broadcasting Corp. (Arabic), Dubai TV (Arabic), China Central TV (Chinese) and New Tang Dynasty Television (Chinese). The web text was collected from Google and Yahoo newsgroups.

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 ]

Ilya Ahtaridis

Membership Coordinator

--------------------------------------------------------------------

Linguistic Data Consortium                  Phone: 1 (215) 573-1275

University of Pennsylvania                    Fax: 1 (215) 573-2175

3600 Market St., Suite 810                        ldc@ldc.upenn.edu

Philadelphia, PA 19104 USA                 http://www.ldc.upenn.edu