Membership year 2027 publication preview
Fall 2026 data scholarship recipients
New publications:
Global TIMIT L1 Tianjin – L2 English
LoReHLT Turkish Representative Language Pack
Multi-Language Conversational Telephone Speech 2014 – French Group
_______________________________________________________________________
Membership year 2027 publication preview
The 2027 membership year is approaching and plans for next year’s publications are in progress. Among the expected releases are:
- Fisher English Training Part 3 Speech: 1300+ hours of previously unreleased English conversational telephone speech from the Fisher collection with speaker metadata
- Computational Cultural Understanding: multimodal data and annotations, assessments, results and supplemental resources to support system development for analyzing and interpreting sociocultural norms, recognizing emotion across cultural contexts, and detecting changes that affect cultural understanding; includes Chinese, Japanese, Russian, Spanish and Turkish releases
- Urdu Conversational Telephone Speech and Transcripts: 74 Urdu telephone conversations (14 hours) with transcripts, developed to support language identification, spoken language recognition and transcription of the Urdu language
- Speech Biomarkers: 23 hours of spontaneous and elicited speech and corresponding transcripts from American English speakers designed to establish a baseline of language and speech patterns from healthy speakers for comparison with clinical populations such as persons with dementia
- Global TIMIT: a continuing series of TIMIT-like read speech corpora, focusing on Mandarin Chinese, Mandarin Chinese dialects and English
Fall 2026 data scholarship recipients
Congratulations to the recipients of LDC's Fall 2026 data scholarships:
Muhammad Azeem: Comenius University Bratislava; Mathematical Institute, Slovak Academy of Sciences (Slovakia): PhD candidate. Muhammad is awarded copies of 2008 NIST Speaker Recognition Evaluation Training Set Part 1 LDC2011S05 and UCLA Speaker Variability Database LDC2021S09 for his work in speaker recognition and pairwise classification.
Cole Baisch: Claremont Graduate University (USA): PhD candidate, Positive Development Psychology, Division of Behavioral and Organizational Sciences. Cole is awarded copies of Switchboard-1 Release 2 LDC97S62 and NXT Switchboard Annotations LDC2009T26 for his work in conversational quality.
Chungche Huang: National Taiwan University (Taiwan): PhD candidate, Graduate Institute of Linguistics. Chungche is awarded a copy of ETS Corpus of Non-Native Written English LDC2014T06 for his work in LLM-based automated essay scoring.
Zhengyang Li, PhD candidate, and Miguel von Aspern, M. Sc.: Technical University of Braunschweig (Germany): Institute of Communications Technology. Zhengyang and Miguel are awarded copies of Switchboard-1 Release 2 LDC97S62, Fisher English Training Speech, Part 1 Speech LDC2004S13, Fisher English Training Speech, Part 1 Transcripts LDC2004T19, Fisher English Training Part 2, Speech LDC2005S13 and Fisher English Training, Part 2 Transcripts LDC2005T19 for their work in low-latency and full-duplex spoken language modeling.
Garrett Oliver: Presbyterian College (USA): B.A. English candidate, English Department. Garrett is awarded a copy of Penn Parsed Corpora of Historical English Second Edition LDC2025T09 for his work in Celtic influence in English.
New publications:
Global TIMIT L1 Tianjin – L2 English was developed by LDC and Tianjin University and contains 43 hours of read speech and transcripts in the Tianjin dialect of Mandarin Chinese and English. Fifty speakers read sentences from an everyday Chinese sentence list, Chinese Gigaword Fifth Edition (LDC2011T13), TIMIT Acoustic-Phonetic Continuous Speech Corpus (LDC93S1) and Treebank-3 (LDC99T42). Some sentences were read by all speakers, others were read by a few speakers, and others were read by just one speaker.
The corpus was collected at Tianjin University, China. Speakers (25 female, 25 male) were born and raised in one of the six urban districts in Tianjin, China and spoke Tianjin. They were also fluent in English based on passing scores on a standard English assessment test.
All recordings were aligned into word, phone and tone segments. Each recording file has corresponding phone, tone, and word segmentation files and Praat TextGrid files.
The Global TIMIT project aimed to create a series of corpora in a variety of languages with a similar set of key features as in the original TIMIT Acoustic-Phonetic Continuous Speech Corpus (LDC93S1) which was designed for acoustic-phonetic studies and for the development and evaluation of automatic speech recognition systems.
2026 members can access this corpus through their LDC accounts. Non-members may license this data for a fee.
LoReHLT Turkish Representative Language Pack was developed by LDC and contains 422 million words of Turkish monolingual text, about 830,000 words of which were translated into English, 84,000 Turkish words translated from English data, and 25 hours of Turkish broadcast news and amateur web audio recordings. Approximately 88,000 words were annotated for named entities and over 27,000 words were annotated for full entity including nominals and pronouns. Noun-phrase chunking was applied to more than 9,300 words. Over 6,200 words were morphologically aligned, and more than 25,400 words were labeled with simple semantic annotation. Topic annotation was applied to the audio recordings. Data was collected from discussion forum, news, reference, social network, broadcast news, web audio recordings, and weblogs.
LoReHLT was a companion project of the DARPA LORELEI program. The LORELEI (Low Resource Languages for Emergent Incidents) program was concerned with building human language technology for low resource languages in the context of emergent situations. Representative languages were selected to provide broad typological coverage.
2026 members can access this corpus through their LDC accounts. Non-members may license this data for a fee.
Multi-Language Conversational Telephone Speech 2014 – French Group was developed by LDC and contains 156 hours of Haitian Creole and West African French telephone speech. The data was collected to support research and technology evaluation in automatic language identification; portions of these recordings were used in the NIST 2015 and 2017 language recognition evaluations. The collection focused on language pair discrimination for 20 languages/dialects, some of which could be considered mutually intelligible or closely related.
This corpus contains 722 recordings: 412 Haitian Creole recordings and 310 West African French recordings. Participants were recruited by native speakers who contacted acquaintances in their social network. Those native speakers made one call, up to 8 minutes, to each acquaintance. Human auditors labeled the calls for language, callee gender, native speaker status and speech clarity.
2026 members can access this corpus through their LDC accounts. Non-members may license this data for a fee.
No comments:
Post a Comment