Monday, June 17, 2019

LDC 2019 June Newsletter

In this newsletter:

New Publications:
USC-SFI MALACH Interviews and Transcripts English – Speech Recognition Edition
First DIHARD Challenge Development - Eight Sources
_____________________________________________________________________

New publications:  

(1) DEFT Spanish Committed Belief Annotation was developed by LDC and consists of approximately 67,000 tokens of Spanish discussion forum text annotated for "committed belief," which marks the level of commitment displayed by the author to the truth of the propositions expressed in the text.

DARPA's Deep Exploration and Filtering of Text (DEFT) program aimed to address remaining capability gaps in state-of-the-art natural language processing technologies related to inference, causal relationships and anomaly detection. LDC supported the DEFT program by collecting, creating and annotating a variety of data sources. 

DEFT Spanish Committed Belief Annotation is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a cost. 

*

(2) USC-SFI MALACH Interviews and Transcripts English – Speech Recognition Edition was developed by IBM as part of the MALACH (Multilingual Access to Large Spoken ArCHives) Project and contains approximately 168 hours of interviews from 682 Holocaust witnesses along with transcripts, a lexicon and other documentation. This release augments USC-SFI MALACH Interviews and Transcripts English (LDC2012S05) by modifying and updating a subset of the original corpus for use with speech recognition systems, such as the Kaldi toolkit.

Specifically, the audio data has been converted from unsegmented mpeg files to a segmented flac compressed format. The speaker-turn, time-stamped transcripts have been updated to an utterance-by-utterance format. A lexicon mapping words to phonemes is provided, and the data is divided into development and training sets.  

The goal of the MALACH project was to develop methods for improved access to large multinational spoken archives in order to advance the state of the art of automatic speech recognition and information retrieval. The characteristics of the USC-SFI collection -- unconstrained, natural speech filled with disfluencies, heavy accents, age-related coarticulations, un-cued speaker and language switching and emotional speech -- were considered well-suited for that task. 

USC-SFI MALACH Interviews and Transcripts English – Speech Recognition Edition is distributed via web download. 

2019 Subscription Members will automatically receive copies of this corpus provided they have submitted a completed copy of the special license agreement. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data at no cost.

*

(3) First DIHARD Challenge Development - Eight Sources was developed by LDC and contains approximately 17 hours of English and Chinese speech data along with corresponding annotations used in support of the First DIHARD Challenge. This release, when combined with First DIHARD Challenge Development - SEEDLingS (LDC2019S10), contains the development set audio data and annotation (diarization, segmentation) as well as the official scoring tool.


The First DIHARD Challenge was an attempt to reinvigorate work on diarization through a shared task focusing on "hard" diarization; that is, speech diarization for challenging corpora where there was an expectation that existing state-of-the-art systems would fare poorly. As such, it included speech from a wide sampling of domains representing diversity in number of speakers, speaker demographics, interaction style, recording quality, and environmental conditions as follows (all sources are in English unless otherwise indicated):
  • Autism Diagnostic Observation Schedule (ADOS) interviews
  • DCIEM/HCRC map task (LDC96S38)
  • Audiobook recordings from LibriVox
  • Meeting speech from 2004 Spring NIST Rich Transcription (RT-04S) Development (LDC2007S11) and Evaluation (LDC2007S12) releases.
  • 2001 U.S. Supreme Court oral arguments
  • Sociolinguistic interviews from SLX Corpus of Classic Sociolinguistic Interviews (LDC2003T15)
  • Chinese video collected by LDC as part of the Video Annotation for Speech Technologies (VAST) project
  • YouthPoint radio interviews

First DIHARD Challenge Development - Eight Sources is distributed via web download. 

2019 Subscription Members will automatically receive copies of this corpus.  2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a cost.

*

(4) First DIHARD Challenge Development - SEEDLingS was developed by Duke University and LDC and contains approximately two hours of English child language recordings along with corresponding annotations used in support of the First DIHARD Challenge. This release, when combined with First DIHARD Challenge Development - Eight Sources (LDC2019S09), contains the development set audio data and annotation (diarization, segmentation) as well as the official scoring tool.

The source data was drawn from the SEEDLingS (The Study of Environmental Effects on Developing Linguistic Skills) corpus, designed to investigate how infants' early linguistic and environmental input plays a role in their learning. Recordings for SEEDLingS were generated in the home environment of 44 infants from 6-18 months of age in the Rochester, New York area. A subset of that data was annotated by LDC for use in the First DIHARD Challenge.

The First DIHARD Challenge was an attempt to reinvigorate work on diarization through a shared task focusing on "hard" diarization; that is, speech diarization for challenging corpora where there was an expectation that existing state-of-the-art systems would fare poorly. As such, it included speech from a wide sampling of domains representing diversity in number of speakers, speaker demographics, interaction style, recording quality, and environmental conditions.

First DIHARD Challenge Development – SEEDLingS is distributed via web download.

2019 Subscription Members will receive copies of this corpus provided they have submitted a completed copy of the special license agreement. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a cost.

*

Monday, May 20, 2019

LDC 2019 May Newsletter

New Publications:
Multi-Language Conversational Telephone Speech 2011 -- English Group  
IARPA Babel Guarani Language Pack IARPA-babel305b-v1.0c __________________________________________________________ 

New publications: 

(1) Multi-Language Conversational Telephone Speech 2011 -- English Group was developed by LDC and is comprised of approximately 18 hours of telephone speech in two general varieties of English: American and South Asian. 

The data were collected primarily to support research and technology evaluation in automatic language identification, and portions of these telephone calls were used in the NIST 2011 Language Recognition Evaluation (LRE). Participants were recruited by native speakers who contacted acquaintances in their social network. Those native speakers made one call, up to 15 minutes, to each acquaintance. Calls are labeled by human auditors for callee gender, dialect type, and noise.   

LDC has also released the following as part of the Multi-Language Conversational Telephone Speech 2011 series: 

Multi-Language Conversational Telephone Speech 2011 -- English Group is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.
*
(2) TAC KBP Chinese Regular Slot Filling - Comprehensive Training and Evaluation Data 2014 was developed by LDC and contains training and evaluation data produced in support of the TAC KBP Chinese Regular Slot Filling evaluation track conducted in 2014. This release includes queries, the 'manual runs' (human-produced responses to the queries), the final rounds of assessment results, and the complete set of Chinese source documents.

The regular Chinese Slot Filling evaluation track involved mining information about entities from text. In completing the task, participating systems and LDC annotators searched a corpus for information on certain attributes (slots) of person and organization entities and attempted to return all valid answers (slot fillers) in the source collection.

TAC KBP Chinese Regular Slot Filling - Comprehensive Training and Evaluation Data 2014 is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.

*
(3) CIEMPIESS Experimentation (Corpus de Investigación en Español de México del Posgrado de Ingeniería Eléctrica y Servicio Social) was developed by the Facultad de Ingeniería at the National Autonomous University of Mexico (UNAM) and consists of approximately 22 hours of Mexican Spanish broadcast and read speech with associated transcripts. The goal of this work was to create acoustic models for automatic speech recognition. For more information and documentation see the CIEMPIESS-UNAM Project website.

CIEMPIESS Experimentation is a set of three different data sets, specifically Complementary, Fem, and Test. Complementary is a phonetically-balanced corpus of isolated Spanish words spoken in Central Mexico. Fem contains broadcast speech from 21 female speakers, collected to balance by gender the number of recordings from male speakers in other CIEMPIESS collections. Test consists of 10 hours of broadcast speech and transcripts and is intended for use as a standard test data set alongside other CIEMPIESS corpora.

Most of the speech recordings in Fem and Test were collected from Radio-IUS, a UNAM radio station. Other recordings were taken from IUS Canal Multimedia and Centro Universitario de Estudios Jurídicos (CUEJ UNAM). Those two channels feature videos with speech around legal issues and topics related to UNAM. The Complementary recordings consist of read speech collected for that corpus.

LDC has released the following data sets in the CIEMPIESS series: 
CIEMPIESS Experimentation is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data at no cost.

*
(4) IARPA Babel Guarani Language Pack IARPA-babel305b-v1.0c was developed by Appen for the IARPA (Intelligence Advanced Research Projects Activity) Babel program. This corpus contains approximately 198 hours of Guarani conversational and scripted telephone speech collected in 2014 and 2015 along with corresponding transcripts.

The Guarani speech in this release represents that spoken in Paraguay. The gender distribution among speakers is approximately equal; speakers' ages range from 16 years to 67 years. Calls were made using different telephones (e.g., mobile, landline) from a variety of environments including the street, a home or office, a public place, and inside a vehicle.

IARPA Babel Guarani Language Pack IARPA-babel305b-v1.0c is distributed via web download.

2019 Subscription Members will receive copies of this corpus provided they have submitted a completed copy of the special license agreement. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.
*

Monday, April 15, 2019

LDC 2019 April Newsletter

LDC at ICASSP 2019

LDC data and commercial technology development


New Publications:
BOLT Egyptian-English Word Alignment -- Discussion Forum Training
Chinese Abstract Meaning Representation 1.0
HAVIC MED Progress Test -- Videos, Metadata and Annotation ____________________________________________________________

LDC at ICASSP 2019
LDC will be exhibiting at ICASSP 2019, held this year May 12-17 in Brighton, UK. Stop by booth 5 to learn more about recent developments at the Consortium and new publications.

LDC will post conference updates via our Twitter feed and Facebook page. We hope to see you there!

LDC data and commercial technology development

For-profit organizations are reminded that an LDC membership is a pre-requisite for obtaining a commercial license to almost all LDC databases. Non-member organizations, including non-member for-profit organizations, cannot use LDC data to develop or test products for commercialization, nor can they use LDC data in any commercial product or for any commercial purpose. LDC data users should consult corpus-specific license agreements for limitations on the use of certain corpora. Visit the Licensing page for further information.

New publications:

(1) BOLT Egyptian-English Word Alignment -- Discussion Forum Training was developed by LDC and consists of 400,448 words of Egyptian Arabic and English parallel text enhanced with linguistic tags to indicate word relations.

The source data in this release consists of discussion forum threads harvested from the Internet by LDC using a combination of manual and automatic processes and is released as BOLT Arabic Discussion Forums (LDC2018T10).

The BOLT word alignment task was built on treebank annotation. Egyptian source tree tokens for word alignment were automatically extracted from tree files of BOLT Egyptian Arabic Treebank annotation on the discussion forum data. Human annotators then followed LDC guidelines to link words and phrases in Arabic to those in English.

BOLT Egyptian-English Word Alignment -- Discussion Forum Training is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.


*

(2) Chinese Abstract Meaning Representation 1.0 was developed by Brandeis University and Nanjing Normal University and is comprised of semantic representations of a set of Chinese sentences from the weblog and discussion forum portions of Chinese Treebank 8.0 (LDC2013T21). Annotations were applied to 10,149 sentences, with 176 sentences unannotated.

Abstract Meaning Representation (AMR) captures "who is doing what to whom" in a sentence. Each sentence is paired with a graph that represents its whole-sentence meaning in a tree structure. Chinese AMR is based on the annotation methodology developed for English with adaptations for handling specific Chinese phenomena. The goal of the Chinese AMR project is to create a large aligned AMR corpus, of which this data set is the first release. For more information about the project, see the Chinese AMR homepage.

Chinese Abstract Meaning Representation 1.0 is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.


*

(3) HAVIC MED Progress Test -- Videos, Metadata and Annotation was developed by LDC and is comprised of approximately 3,650 hours of user-generated videos with annotation and metadata.

In a collaboration with NIST (the National Institute of Standards and Technology) to advance multimodal event detection and related technologies, LDC developed a large, heterogeneous, annotated multimodal corpus for HAVIC (the Heterogeneous Audio Visual Internet Collection) that was used in the NIST-sponsored MED (Multimedia Event Detection) task for several years. HAVIC MED Progress Test is a subset of that corpus, specifically, a collection of event and background videos originally released to support the 2012-2015 MED tasks.

This release consists of videos of various events (event videos) and videos completely unrelated to events (background videos) harvested by a large team of human annotators. Each event video was manually annotated with a set of judgments describing its event properties and other salient features. Background videos were labeled with topic and genre categories.

HAVIC MED Progress Test -- Videos, Metadata and Annotation is distributed via hard drive.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. This corpus is a members-only release and is not available for non-member licensing. Contact ldc@ldc.upenn.edu for information about membership.

Friday, March 15, 2019

LDC 2019 March Newsletter

Call for Papers - LTC 2019, LREC 2020

New Publications:
___________________________________________________________

Call for Papers

The 9th Language & Technology Conference (LTC 2019) will take place on May 17-19, 2019 at the Adam Mickiewicz University in Poznań, Poland. LTC addresses Human Language Technologies as a challenge for computer science, linguistics and related fields. Conference papers are due next week on Wednesday, March 20, 2019 (midnight, any time zone). For more information, visit the conference webpage

The 12th Conference on Language Resources and Evaluation (LREC 2020) will take place on May 13-15, 2020 at the Palais du Pharo in Marseille, France. LREC aims to provide an overview of the state-of-the-art, explore new R&D directions and emerging trends, and exchange information regarding language resources and their applications, evaluation methodologies and tools. Conference papers are due by November 25, 2019. For more information, including conference topics, visit the conference webpage.

New Publications:

(1) CALLFRIEND Egyptian Arabic Second Edition was developed by LDC and consists of approximately 25 hours of unscripted telephone conversations between native speakers of Egyptian Arabic. This second edition updates the audio files to wav format, simplifies the directory structure and adds documentation and metadata. The first edition is available as CALLFRIEND Egyptian Arabic (LDC96S49).

All data was collected before July 1997. Participants could speak with a person of their choice on any topic; most called family members and friends. All calls originated in North America. The recorded conversations last up to 30 minutes. 

CALLFRIEND Egyptian Arabic Second Edition is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.  

*

(2) Penn Discourse Treebank Version 3.0 is the third release in the Penn Discourse Treebank project, the goal of which is to annotate the Wall Street Journal (WSJ) section of Treebank-2 (LDC95T7) with discourse relations. Penn Discourse Treebank Version 2 (LDC2008T05) contains over 40,600 tokens of annotated relations. In Version 3, an additional 13,000 tokens were annotated, certain pairwise annotations were standardized, new senses were included and the corpus was subject to a series of consistency checks.

This corpus contains two tools: (1) The Annotator, used for annotation and adjudication, and which can also be used for viewing the corpus; and (2) The Conversion Tool for converting Version 2 annotation files into the Version 3 format.

The documentation directory contains a manual describing what is new in Version 3 and how Version 3 differs from Version 2; the methods and guidelines used in annotating PDTB Version 3; and a range of statistics on the tokens, including the frequency of each connective, its sense labels and its modifiers. More information about the corpus and research carried out by the developers and others using the corpus can be found on the PDTB website.

Penn Discourse Treebank Version 3.0 is distributed via web download. 

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.  

*

(3) VAST Chinese Speech and Transcripts was developed by LDC for the VAST (Video Annotation for Speech Technologies) project and is comprised of approximately 29 hours of Mandarin Chinese audio extracted from amateur video content harvested from the web and corresponding time-aligned transcripts. 

Audio files were transcribed using XTrans, which supports manual transcription across multiple channels, languages and platforms. Transcribers followed a Quick-Rich Transcription style; transcription guidelines are included in this release. 

The aim of the VAST project was to collect and annotate data in several languages to support the development of speech technologies such as speech activity detection, language identification, speaker identification, and speech recognition. 

VAST Chinese Speech and Transcripts is distributed via web download.

2019 Subscription Members will automatically receive copies of this corpus. 2019 Standard Members may request a copy as part of their 16 free membership corpora. Non-members may license this data for a fee.