- Corpora - ELRA lists

November 2022 Newsletter - LDC
by Penn LDC 15 Nov '22

15 Nov '22

In this newsletter: Join LDC for membership year 2023 Fall 2022 data scholarship recipients Spring 2023 data scholarship application deadline 30th Anniversary Highlight: CALLFRIEND New publications: BOLT English Translation Treebank - Egyptian Arabic SMS/Chat<https://catalog.ldc.upenn.edu/LDC2022T06> Samrómur Children Icelandic Speech 1.0<https://catalog.ldc.upenn.edu/LDC2022S11> Third DIHARD Challenge Development<https://catalog.ldc.upenn.edu/LDC2022S12> ________________________________ Join LDC for membership year 2023 It's time to renew your LDC membership for 2023. Current (2022) members who renew their membership before March 1, 2023 will receive a 10% discount. New or returning organizations will receive a 5% discount if they join the Consortium by March 1. In addition to receiving new publications, current LDC members enjoy the benefit of licensing older data from our Catalog of 900+ holdings at reduced fees. Current-year for-profit members may use most data for commercial applications. Plans for 2023 publications are in progress. Among the expected releases are: * AIDA Ukrainian Broadcast and Telephone Speech Audio and Transcripts: 156 hours of Ukrainian conversational telephone speech and broadcast news with 1.2 million words of corresponding orthographic transcripts * 2019 NIST SRE: audiovisual and leaderboard challenge sets based on amateur videos and Tunisian Arabic telephone speech, respectively * DEFT English ERE: English text from assorted genres annotated for entities, relations and events * Mixer 3 and Mixer 7 speech collections: thousands of hours of telephone speech and metadata from Mixer 3 (multiple languages) and Mixer 7 (Spanish, plus interviews and transcript readings) * CALLFRIEND Russian: 100 telephone conversations among native speakers, transcripts and a lexicon, released in separate speech and text data sets * REMIX Telephone Collection: English telephone speech from 385 participants in previous Mixer studies * LORELEI: representative and incident language packs containing monolingual text, bi-text, translations, annotations, supplemental resources and related tools in various languages (e.g., Indonesian, Swahili, Tagalog, Tamil, Zulu) For full descriptions of all LDC data sets, browse our Catalog<https://catalog.ldc.upenn.edu/>. Visit Join LDC<https://www.ldc.upenn.edu/members/join-ldc> for details on membership, user accounts and payment. Fall 2022 LDC data scholarship recipients LDC congratulates the following Fall 2022 data scholarship recipients: * Nelson Filipe Costa: Concordia University (Canada); PhD, Machine Learning. Nelson is awarded a copy of Penn Discourse Treebank Version 3.0 (LDC2019T05<https://catalog.ldc.upenn.edu/LDC2019T05>) for his work in discourse relationships and mapping. * Paul Pope: University of Eastern Finland (Finland); MA, Linguistic Data Sciences. Paul is awarded a copy of ETS Corpus of Non-Native Written English (LDC2014T06<https://catalog.ldc.upenn.edu/LDC2014T06>) for his research on text classification. * Abhinav Singh: Sharda University (India); PhD, Forensic Science. Abhinav is awarded a copy of TIMIT Acoustic-Phonetic Continuous Speech Corpus (LDC93S1<https://catalog.ldc.upenn.edu/LDC93S1>) for his research on forensic speech recognition. * Lucas Zheng: Deerfield Academy (USA); High School Scholar. Lucas is awarded copies of Arabic Treebank Part 1 v. 4.1 (LDC2010T13<https://catalog.ldc.upenn.edu/LDC2010T13>) and Arabic Treebank Part 2 v. 3.1 (LDC2011T09<https://catalog.ldc.upenn.edu/LDC2011T09>) for his work on analyzing syntactic and lexical similarities across MSA genres and POS-tagging for MSA. Students can learn more about the LDC data scholarship program on the Data Scholarships page.<https://www.ldc.upenn.edu/language-resources/data/data-scholarships> Spring 2023 data scholarship application deadline Applications are now being accepted through January 15, 2023 for the Spring 2023 LDC data scholarship program which provides university students with no-cost access to LDC data. Consult the LDC Data Scholarship<https://www.ldc.upenn.edu/language-resources/data/data-scholarships>s page for more information about program rules and submission requirements. 30th Anniversary Highlight: CALLFRIEND The CALLFRIEND series is a multi-language collection of unscripted telephone conversations conducted by LDC in the 1990s to support language identification technology development (Liberman & Cieri, 1998<https://www.ldc.upenn.edu/sites/www.ldc.upenn.edu/files/lrec1998-creation-d…>). Covered languages are American English, Canadian French, Egyptian Arabic, Farsi, German, Hindi, Japanese, Korean, Mandarin Chinese, Spanish, Tamil and Vietnamese. For English, Mandarin and Spanish, the collection includes two distinct dialects. Participants could speak with a person of their choice on any topic; most called family members and friends. All calls originated in North America. This speech data was the foundation for NIST's Language Recognition Evaluations<https://www.nist.gov/itl/iad/mig/language-recognition> conducted from 1996-2007. The first editions of the CALLFRIEND series published in LDC's Catalog in 1996 contain 60 calls evenly split into 20 calls each for a training partition to develop language models, a development partition for parameter tuning, and an evaluation partition to test performance (Torres-Carrasquillo, et al., 2004<https://www.isca-speech.org/archive_open/archive_papers/odyssey_04/ody4_297…>). Beginning in 2014, LDC released second editions for American English (LDC2019S21<https://catalog.ldc.upenn.edu/LDC2019S21>, LDC2020S08<https://catalog.ldc.upenn.edu/LDC2020S08>), Canadian French (LDC2019S18<https://catalog.ldc.upenn.edu/LDC2019S18>), Egyptian Arabic (LDC2019S04<https://catalog.ldc.upenn.edu/LDC2019S04>), Farsi (LDC2014S01<https://catalog.ldc.upenn.edu/LDC2014S01>), and Mandarin Chinese (LDC2018S09<https://catalog.ldc.upenn.edu/LDC2018S09>, LDC2020S06<https://catalog.ldc.upenn.edu/LDC2020S06>). The goal of the second editions is to facilitate continued widespread use of the data, specifically, by updating the audio files to .wav format, simplifying the directory structure, adding documentation and metadata, and combining the training, development and evaluation splits. CALLFRIEND Farsi Second Edition also includes additional telephone recordings and a separate transcripts release (LDC2014T01<https://catalog.ldc.upenn.edu/LDC2014T01>). In addition to work on language identification, CALLFRIEND corpora have been used in a variety of research tasks, including subject omission in Korean (Lee 2012<https://www.dbpia.co.kr/journal/articleDetail?nodeId=NODE10893918>), contemporary Persian vowels in casual speech (Jones 2019<https://repository.upenn.edu/pwpl/vol25/iss1/15/>), Mandarin telephone closings among familiars (Huang, 2020<https://www.jbe-platform.com/content/journals/10.1075/ap.19017.hua>), and adjective constructions in English conversation (Bybee & Thompson, 2021<https://www.mdpi.com/2226-471X/7/1/2/htm>), among many others. To learn more about the CALLFRIEND collection or about other LDC corpora used for language identification research, search the Catalog<https://catalog.ldc.upenn.edu/search> by the "recommended application" and select "language identification" from the list. ________________________________ New publications: BOLT English Translation Treebank - Egyptian Arabic SMS/Chat<https://catalog.ldc.upenn.edu/LDC2022T06> was developed by LDC and consists of SMS and chat text data (472 files representing 98,206 tokens) translated from Egyptian Arabic to English and annotated for part-of-speech and syntactic structure. Only the translated English text is included in the source data for this release. Part-of-speech and treebank annotation conformed to Penn Treebank II style, incorporating changes to those guidelines that were developed under the GALE (Global Autonomous Language Exploitation) program. Supplementary guidelines for English treebanks and web text are included in the corpus documentation. 2022 members can access this corpus through their LDC accounts. Non-members may license this data for a fee. * Samrómur Children Icelandic Speech 1.0<https://catalog.ldc.upenn.edu/LDC2022S11> was developed by the Language and Voice Lab, Reykjavik University<https://lvl.ru.is/> in cooperation with Almannarómur, Center for Language Technology<https://almannaromur.is/>. The corpus contains 131 hours of Icelandic prompted speech from 3,175 speakers (children, aged 4-17 years) representing 137,597 utterances. Speech data was collected between October 2019 and September 2021 using the Samrómur website<https://samromur.is> which displayed prompts to participants. The prompts were mainly from The Icelandic Gigaword Corpus<http://clarin.is/en/resources/gigaword>, which includes text from novels, news, plays, and from a list of location names in Iceland. Additional prompts were taken from the Icelandic Web of Science<https://www.visindavefur.is/> and others were created by combining a name followed by a question or a demand. Prompts and speaker metadata are included in the corpus 2022 members can access this corpus through their LDC accounts provided they have submitted a completed copy of the special license agreement. Non-members may license this data for a fee. * Third DIHARD Challenge Development<https://catalog.ldc.upenn.edu/LDC2022S12> was developed by LDC and contains approximately 34 hours of English and Chinese speech data along with corresponding annotations used in support of the Third DIHARD Challenge<https://dihardchallenge.github.io/dihard3>. The DIHARD third development and evaluation sets were drawn from diverse sources including monologues, map task dialogues, broadcast interviews, sociolinguistic interviews, meeting speech, speech in restaurants, clinical recordings, and amateur web videos. Annotations include diarization and segmentation. 2022 members can access this corpus through their LDC accounts. Non-members may license this data for a fee. To unsubscribe from this newsletter, log in to your LDC account<https://catalog.ldc.upenn.edu/login> and uncheck the box next to "Receive Newsletter" under Account Options; or contact LDC for assistance.

1 0

Research Assistant for Voice- and Chatbots at TU Berlin
by Roland Roller 15 Nov '22

15 Nov '22

The Quality and Usability Group at Technical University Berlin is searching for a research assistant working on spoken and text-based interactive dialog-systems. Further information about the position can be found here: https://tub.stellenticket.de/en/offers/157021/ If you have got questions about the position, please contact my colleague Stefan Hillman (stefan.hillmann(a)tu-berlin.de). The closing date is the 5th of December 2022.

1 0

CfP ICAME44
by Bertus van Rooy 15 Nov '22

15 Nov '22

The 44th annual conference of the International Computer Archive of Modern and Medieval English (ICAME44) is hosted by the North-West University in Vanderbijlpark, South Africa, from 17 to 21 May 2023. The conference format will be hybrid, with on-site and on-line participation possible. The conference theme is “English going places, Corpora crossing spaces”. Participants are invited to reflect on how the diffusion of English across spaces – physical, virtual or metaphorical – contributes to shape the language. Confirmed keynote speakers are: Alexandra Esimaje (Benson Idahosa University, Nigeria) “Challenges of doing corpus linguistics in Africa” Haidee Kotze (Utrecht University, The Netherlands) “Protean invisibilities: A corpus-based perspective on the role of translation in language variation and change in global Englishes” Christiane Meierkord (Ruhr University Bochum, Germany) “Corpora crossing disciplinary borders: ICE Uganda meets sociolinguistics, anthropology and ethnography” Mike Scott (Lexical Analysis Software & Aston University, United Kingdom) “News Downloads and Text Coverage: Case Studies in Relevance” Submission Guidelines All papers must be original and not simultaneously submitted to another journal or conference. The following paper categories are welcome: Full papers (20 minutes presentation,10 minutes discussion) describing the findings of new research projects involving the study of the English language by means of corpora. Work-in-Progress reports (10 minutes presentation, 5 minutes discussion) describing ongoing projects prior to final results. Posters describing new ideas or new findings. When submitting your abstract, please indicate in the title line whether you wish to be considered for a full paper, WiP report or poster. Please submit your abstract via EasyChair at https://easychair.org/conferences/?conf=icame44 The submission deadline for early review (before Christmas) is 2 December 2022, and otherwise 23 December 2022 for review done by end of January 2023. Venue The conference will be held in Vanderbijlpark, South Africa, at the Emerald Resort. The resort simultaneously provides the venue and accommodation. It is about 95km away from OR Tambo International Airport in Johannesburg. Transport from the airport be arranged to conference attendees who require it on the two days prior to the conference, and back to the airport after the completion of the conference, as there are no direct public transport options linking the airport and Vanderbijlpark. There is a direct connection by highway, for attendees who want to rent a vehicle or book a taxi/Uber in the alternative. Contact All questions about submissions or the conference should be emailed to the conference organiser, Bertus van Rooy, using e-mail address a.j.vanrooy [at] uva.nl. More information is also available on the conference website - ICAME44 | humanities.nwu.ac.za - but it is continuously being updated with further information.

1 0

Call-for-workshop-proposals: ACM ICMR 2023
by Bogdan Ionescu 15 Nov '22

15 Nov '22

[Apologies for multiple postings] ACM International Conference on Multimedia Retrieval 2023 Thessaloniki, Greece, 12-15 June 2023 https://icmr2023.org/ *** Call for Workshop Proposals *** The ACM International Conference on Multimedia Retrieval in 2023 (ICMR 2023) is calling for high quality Workshops addressing innovative research in multimedia retrieval and its related broad fields. The scope of the conference is not limited to search and retrieval of multimedia data but also addresses analysis and understanding of multimedia contents, including: - Integration of diverse multimodal data. - Deep learning-based methodology and practical multimedia applications. - Multimedia dataset creation and management (and compliance with ethical practices). - Interpretability of multimedia analysis and retrieval techniques and connection with social sciences. - Exploring bias in multimedia datasets. - Practical systems/applications of multimedia analysis and retrieval on different areas, e.g., media & entertainment, culture, health, security, etc. - Multimedia analysis and retrieval aiming to solve major societal challenges (e.g., disinformation, climate change, social inequalities etc.). - Legal/ethical challenges and societal impact of multimedia analysis and retrieval. - Community-contributed social data. - Lifelogging data and automatically generated sensor data. The Workshops will contain both: - Long research papers which should present complete work with evaluations on topics related to the Conference. They will have both oral and poster presentations at the conference. - Short research papers which should present preliminary results or more focused contributions. They will be presented as posters at the conference. *** Important Dates *** Submit your workshop proposal until December 15, 2022. *** Submission Instructions *** Please, download the Workshop Proposal template from the conference website. Please, send your proposal as a pdf file to the Workshops Chairs at icmr2023-workshops(a)iti.gr. *** Contact *** For any question regarding submissions for workshop proposals, please visit the conference website or email the Workshops Chairs (icmr2023-workshops(a)iti.gr). *** Workshops Chairs *** Dr. Ichiro Ide (Nagoya University) Prof. Lucile Sassatelli (Université Côte d’Azur) Prof. Ioannis (Yiannis) Patras (Queen Mary University of London) Dr. Filareti Tsalakanidou (Centre for Research and Technology Hellas) On behalf of the organizers, Bogdan Ionescu https://www.aimultimedialab.ro/

1 0

Erinnerung: Förderung einer Open-Access-Publikation „Einführung in die Computerlinguistik“ (Deadline 15.12.2022)
by josef Ruppenhofer 14 Nov '22

14 Nov '22

(Apologies if you end up receiving this multiple times. The text below appears only in German as it concerns a call for a German language book proposal.) Aufforderung zur Konzepteinreichung Zum Hintergrund Das letzte Jahrzehnt hat weltweit zahlreiche neue Aktivitäten und Entwicklungen auf dem Feld der Computerlinguistik erbracht. Nicht nur haben sich neue Ansätze und Methoden durchgesetzt; die Computerlinguistik als solche hat sich auch disziplinär stark ausdifferenziert. Dessen ungeachtet ist festzustellen, dass für Forschung und Lehre kein aktuelles deutschsprachiges Lehrbuch zur Verfügung steht, das ein umfassendes Bild des aktuellen Standes der Computerlinguistik zu zeichnen vermag. Die GSCL möchte deshalb dazu beitragen, diese Lücke zu füllen und hat beschlossen, Mittel zur Veröffentlichung eines umfassenden einführenden Lehrbuches oder einander ergänzender Lehrbücher zur Verfügung stellen. Was wird gefördert? Die Einführung soll sich vorrangig an den wissenschaftlichen Nachwuchs, insbesondere an Studierende in einem Bachelor-Studiengang, richten. Diese allgemeine Einführung in den Forschungsgegenstand sollte zudem die Trends der letzten Jahre thematisieren. Der Fokus der Publikation oder der Publikationen könnte beispielsweise auf anwendungsorientierte Fragestellungen der Informatik gelegt werden. Alternativ könnten auch Schwerpunktsetzungen in den Bereichen Linguistik oder Digital Humanities erfolgen. Die GSCL wird die finanziellen Mittel bereitstellen, die zur Deckung der Publikationskosten eines oder mehrerer sich ergänzender deutschsprachiger Open- Access-Bände zur Einführung in die Computerlinguistik in einem renommierten Verlag benötigt werden. Sofern die Autorinnen und Autoren bzw. Herausgeberinnen und Herausgeber die Publikation bei einem kostenfreien Non-Profit-Verlag bevorzugen, unterstützt die GSCL auch solche Vorhaben, kann die dadurch freiwerdenden finanziellen Mittel allerdings nicht umwidmen. Momentan steht die GSCL mit renommierten Verlagen in Kontakt, die Autorinnen und Autoren nicht nur ein breitgefächertes Distributionsnetz, sondern auch deutliche Mehrwerte für die Sichtbarkeit und Reichweite ihrer Forschung bieten. Die Einführung soll so konzipiert werden, dass sie sowohl in digitaler Fassung als PDF über Bibliotheken und Gedächtnisorganisationen barrierefrei zur Verfügung steht, als auch als Druckversion über den Buchhandel verbreitet und von Bibliotheken angeschafft werden kann. Die GSCL wird alle mit der Publikation anfallenden Kosten von professionellem Lektorat, Satz und dauerhafter Bereitstellung des Bandes übernehmen. Es steht zu erwarten, dass ein Zuschlag für die erfolgreichen Autorinnen und Autoren erhebliche Reputationsgewinne infolge der Schließung einer Lücke in der deutschsprachigen Computerlinguistik zur Folge haben wird. Weitere Anreize, insbesondere Honorare für das Verfassen von Texten oder die Herausgabe des Bandes werden von der GSCL explizit nicht ausgelobt. Zur Antragstellung Die GSCL bittet einschlägig erfahrene Wissenschaftlerinnen und Wissenschaftler um Einreichung eines Essays zur thematischen Ausrichtung einer Einführung. Die Einführung kann sowohl als monographische Schrift als auch als Sammelband konzipiert werden. Bis zum 15.12.2022 sind Autorinnen und Autoren oder Herausgeberinnen und Herausgeber aufgerufen, eine entsprechende kurze Konzeption einzureichen, welche die thematische Ausrichtung des Lehrbuches konzise skizziert und, im Falle von Sammelbänden, Einzelthemen und potenzielle Autorinnen und Autoren aufführt. Erwartet wird überdies ein straffer, aber umsetzbarer Zeitplan zur Durchführung der Publikation, die idealerweise innerhalb eines Jahres abzuschließen sein wird. Die Auswahl der Einreichungen wird bis zum 31.01.2023 erfolgen; entsprechende Vorschläge sind bitte unter dem Betreff „GSCL: Einführung in die Computerlinguistik“ an die Adresse informationsreferent(a)gscl.org zu richten

1 0

Internship offer at INRIA
by Biswesh Mohapatra 14 Nov '22

14 Nov '22

Hello Everyone, I would like to inform you that our team [ http://articulab.hcii.cs.cmu.edu/ | Articulab ] at INRIA Paris is having an internship position for working on understanding pre-trained dialog models. PFA a description of the internship. Ideally, the candidate would start working in January 2023. Warm Regards, Biswesh Mohapatra INRIA

1 0

Call for papers: Explainable AI in Natural Language Processing
by Somnath Banerjee 14 Nov '22

14 Nov '22

*** Apologies for cross-posting *** Call for papers: Explainable AI in Natural Language Processing Traditional Natural Language Processing (NLP) models (e.g., decision trees, Markov models, etc.) have primarily been based on techniques that are inherently interpretable models, referred to as white-box techniques. However, in recent years, NLP models have employed advanced neural approaches along with language embedding features. Using these advanced approaches, mostly referred to as black-box techniques, the NLP models have yielded state-of-art performance. Nonetheless, the level of interpretability (e.g., how the model arrives at its results) has reduced significantly. This obfuscated interpretability not only lowers the end users’ trust in the NLP models but also makes it challenging for the developers to debug or improve by analyzing the models for further improvement. Therefore, nowadays, researchers in the NLP community are giving significant attention to the emerging field called Explainable AI (XAI) to tackle the obfuscated complexity of AI systems for trust and improvement. Apart from academia, organizations and companies also have launched high-funding projects such as DARPA XAI, People +AI Research (PAIR), etc. As XAI is still a growing field, there is plenty of room for innovation to improve the explainability of NLP systems. In recent works, explainable NLP models have captured linguistic knowledge of neural networks, explain predictions, stress-test models via challenge sets or adversarial examples, and interpret language embeddings. The goal of this Research Topic is to better understand the present status of the XAI in NLP by identifying: new dimensions for a better explanation, evaluation techniques used to measure the quality of explanations, approaches or developments of new software toolkits to explain XAI in NLP, and transparent deep learning models for different NLP task. The scope of this Research Topic covers (but is not restricted to) the following topics: • Survey of XAI in NLP in general or any particular NLP task such as NER, QA, Sentiment analysis, social media (SocialNLP), etc. • Explainable Neural models in Machine Translation • Explainable Neural models in Named Entity Recognition • Explainable Neural models in Question Answering • Explainable Neural models in Sentiment Analysis • Explainable Neural models in Opinion Mining • Explainable Neural models in SocialNLP • Evaluation techniques used to measure the quality of explanations • Tools for explaining explainability • Resources related to XAI in the context of NLP The Research Topic welcomes contributions toward interpretable models for efficient solutions to NLP research problems that explain the explainability of the proposed model using suitable explainability technique(s) (e.g., example-driven, provenance, feature importance, induction, surrogate models, etc.), visualization technique(s) (e.g., raw examples, saliency, raw declarative, etc.), and other aspects. Software toolkits or approaches that can help users express explainability to their models and ML pipelines are also welcome. The full Call for Papers is available at https://www.frontiersin.org/research-topics/48440/explainable-ai-in-natural… Impact of the publication: https://www.frontiersin.org/about/impact The current deadlines are: * Abstract Deadline:16 December 2022 - A soft-deadline, thus whilst not mandatory, if you would like feedback on your prospective manuscript’s suitability, I encourage you to submit an abstract around this time. * Manuscript Deadline: 14 April 2023, This is a mandatory deadline for your full manuscript submission. However, we can accommodate personal extensions on a case-by-case basis. Guest Editors: Somnath Banerjee (University of Tartu, somnath.banerjee(a)ut.ee) David Tomás (University of Alicante, dtomas(a)dlsi.ua.es) Somnath Banerjee Lecturer, Institute of Computer Science, University of Tartu, Narva mnt 18, room 3063 51009 Tartu, ESTONIA webpage: http://www.ut.ee//~somnath/

1 0

PhD Thesis defense: Benjamin Muller (Inria Paris, Almanach) 17/11/2022 8h30 "How Can We Make Language Models Better at Handling the Diversity and Variability of Natural Languages?"
by Djamé Seddah 14 Nov '22

14 Nov '22

Hi everyone, I have the pleasure to invite you to my Ph.D. defense next Thursday at 8:30 AM Paris time, it will take place at the Inria Paris Research Centre (2 rue Simone Iff, 75012 Paris) at the Jacques-Louis Lions room. It will be followed by a pot de thèse around 11 am. The Ph.D. committee will be made of: • Yoav Golberg, Bar Ilan University, Rapporteur • ‪Lilja Øvrelid, University of Oslo, Rapportrice • Benjamin Piwowarski, CNRS, Examinateur • Benoît Sagot INRIA Paris, Directeur • Natalie Schulter, IT University of Copenhagen, Examinatrice • Djamé Seddah, INRIA Paris, Superviseur Abstract: Deep Learning techniques applied to Natural Language Processing (NLP) have led to impressive empirical progress in recent years. In essence, this progress is due to the development of better-contextualized representations of textual data that can be easily used --- or transferred --- for a wide variety of NLP tasks. In their most recent and popular forms, these models consist of large-scale deep-learning language models, first pretrained on a large quantity of raw data and then adapted to specific tasks. These language models are now essential for search engines, question-answering pipelines, machine translation systems, etc. However, these models usually require substantial computing power and large amounts of raw textual data. This makes natural language's inherent diversity and variability a vivid challenge in NLP. Indeed, collecting large datasets for low-resource languages is challenging and costly, and training models from scratch for every domain and language is unreasonable in practice. Additionally, understanding the behavior of deep learning-based models is intrinsically tricky, making the development of more cost-effective techniques even more challenging. For these reasons, we focus on the following question: ``How can we make language models better at handling the variability and diversity of natural languages?''. As a starting step, we explore the generalizability of language models by building one of the first large-scale replication of a BERT model for a non-English language. We analyze the critical training ingredients and show that it can achieve state-of-the-art performance with only a few gigabytes of diverse data. Our results raise the question of using these language models on highly-variable domains such as these found in user-generated content. Focusing on domain-gap reduction via lexical normalization, we show that this task can be addressed accurately with BERT-like models. However, we show that it only partially helps downstream performance. In consequence, we focus on direct adaptation techniques using what we refer to as representation transfer and explore challenging settings such as the zero-shot setting, low-resource language varieties like Bambara or Uyghur, and highly variable and non-standardized code-mixed dialects such as a North-African Arabic dialect written in the Latin script. We show that multilingual language models can be adapted and used efficiently with low-resource languages, even with the ones unseen during pretraining, and that the script is a critical component in this adaptation. NLP technologies are becoming increasingly critical to accessing knowledge, connecting with our friends, and extracting meaningful information from large quantities of text. In this thesis, we present concrete and usable solutions to ensure that we can build accurate NLP systems for the most significant number of domains and languages at a reasonable cost. For those who cannot make it in person, a video-conference link will be shared in the following google doc: https://docs.google.com/document/d/1YW1sa_x6oTLiXzgXoc2vcJqxaiZmhyu4TUGgl_D… Best, Benjamin Muller

1 0

LTC 2023 3rd Call for paper : NEW DEADLINE EXTENSION 12th December 2022
by pap 13 Nov '22

13 Nov '22

NEW DEADLINE EXTENSION for paper submission: 12th december 2022 for the 10th LANGUAGE AND TECHNOLOGY CONFERENCE: Human Language Technologies as a Challenge for Computer Science and Linguistics, April 21-23, 2023, Poznań, Poland (LTC 2023) 3rd CALL FOR PAPERS VENUE: April 21-23, 2023 in Poznań, Poland. IMPORTANT DATES/DEADLINES: * Deadline for submission of papers for review: December 12, 2022 * Acceptance/Refusal notification: January 7, 2023 * Deadline for submission of final versions of accepted papers: January 16, 2023 * Conference: April 21-23, 2023 To more info please do visit the WEB SITE: http://ltc.amu.edu.pl/ The Faculty of Mathematics and Computer Science of the Adam Mickiewicz University in Poznań (Poland) and the Adam Mickiewicz University Foundation organize the 10th LTC on April 21-23, 2023. Following the tradition of the past events, it is supported by ELRA, FlaReNet and META-NET. Since 2005 LTC is organized every two years as the “Language & Technology Conference: Human Language Technologies as a Challenge for Computer Science and Linguistics”. Started 27 years ago, the conference find its origin in the Language and Technology Awareness Days, a meeting organized in 1995 with the assistance of the European Commission (DG XIII), with among the key speakers: Antonio Zampolli (Italy), Dafydd Gibbon (Germany), Dan Tufiş (Romania) and Orest Kossak (Ukraina). The conference addresses to researchers, advanced and beginning students, as well as all other people interested in the new advances at the borderline between computer science, human language industries and linguistics. CONFERENCE TOPICS LTC 2023 welcomes the submission of original unpublished papers on various aspects of Human Language Technology of both theoretical and practical importance including: * AI-oriented studies of human language competence * communicative intelligence * computational semantics * computer modeling of language competence * corpora-based methods in language engineering * electronic language resources and tools * formalization of natural languages * HLT related policies * HLT standards and best practices * HLTs as support for e-learning * HLTs as support for foreign language teaching * HLTs as support in solving Homeland Security problems (technology applications and legal aspects) * human-machine NL interfaces * knowledge representation * language-specific computational challenges for HLTs (especially for languages other than English) * legal issues connected with HLTs (problems and challenges) * Logic Programming in Natural Language Processing * methodological issues in HLT * neural networks in language engineering * NL applications in robotics * NL understanding by computers * NL user modeling * NLP methods in cyber-criminality detection and prevention * paralinguistic phenomena in Natural Language Processing * parsing and other forms of NL processing * question answering * sentiment, opinion and emotion analysis * socio-political aspects of HLTs * speech processing * system prototype presentations * technological aspects of nonverbal linguistics * text-based information retrieval and extraction * tools and methodologies for developing multilingual systems * translation enhancement tools * validation in all areas of HLTs * visionary papers in the field of HLT * WordNet-like ontologies PAPER SUBMISSION The conference accepts papers in English only. Papers (5 formatted pages in the conference format) are due by December 12, 2022 (midnight, any time zone) and should not disclose the author(s) in any manner. In order to facilitate submission we have decided to reduce the formatting requirements as much as possible at the submission stage. All submissions are to be made electronically via the LTC 2023 web submission system (EasyChair). Paper templates are available at http://www.ltc.amu.edu.pl (Paper Submission section). PUBLICATION POLICY Acceptance will be based on the reviewers' assessments (blind peer review by tree experts). The accepted papers will be published in the conference proceedings (hard and electronic copy, with ISBN number). The abstracts of the accepted contributions will also be made available via the conference page (during its lifetime). Publication requires full electronic registration and payment of the conference fee (full registration) by at least one co-authors in the due time (dates will be presented at the conference site). One registration fee entitles publication of one paper. Since 2005 until now, post-conference volumes with substantially extended versions of selected conference papers were published. Since 2007 these post-conference monographs were published in the Springer Verlag series Lecture Notes in Artificial Intelligence. We intend to continue this tradition. Papers will be selected based on reviewers’ reports among the best evaluated papers of general interest with new innovative results. Preference will be given to papers providing significant content extension with respect to the paper presented at the conference. LTC’23 SPECIAL TRACKS/WORKSHOPS for LTC Participants will be offered soon (please visit http://www.ltc.amu.edu.pl). AWARDS FOR BEST STUDENT PAPERS Special awards will be granted to the best student (status at the date of paper submission). The majoritarian student contribution to the paper must be confirmed by all paper co-authors in a written form. More details will be provided at the LTC website. To find information out about the winners at previous conferences please visit www.ltc.amu.edu.pl. ACCOMPANYING EVENTS Please do not hesitate to contact us with new suggestions and ideas concerning accompanying events (workshops, exhibits, panels, etc.). Suggestions, ideas and observations may be addressed directly to the LTC Chairs (vetulani(a)amu.edu.pl and pap(a)limsi.fr (cc to ltc23(a)amu.edu.pl and marta.witkowska(a)amu.edu.pl). LANGUAGE: The conference language is English. CONTACTS: Zygmunt Vetulani (vetulani(a)amu.edu.pl) and Patrick Paroubek (pap(a)lisn.fr) (LTC23 Co-chairs), Marta Witkowska (LTC23 Secretary) (marta.witkowska(a)amu.edu.pl) MORE ABOUT THE CONFERENCE at the LTC23 website: http://www.ltc.amu.edu.pl *) City of Poznań: https://www.poznan.pl/mim/en/

1 0

EACL 2023: 2nd Call for System Demonstrations (deadline: Dec 02, 2022)
by Joel Tetreault 11 Nov '22

11 Nov '22

(apologies for cross-posting) ================ EACL 2023 -- Call for System Demonstrations ================== - Website: https://2023.eacl.org/calls/demos/ - Deadline: December 2nd, 2022 - Questions? Email eacl2023demos(a)googlegroups.com The system demonstration track at EACL 2023 is a venue for papers describing system demonstrations, ranging from early prototypes to mature production-ready systems. Publicly available open-source or open-access systems are of special interest. All accepted demos are published in a companion volume of the conference proceedings. We expect at least one of the authors to present a live demo during a demo session at EACL 2023, with an accompanying poster, either virtually or in-person. Submissions will undergo a single-blind reviewing process. Therefore, papers may include author and affiliation information, and freely make references to previously published material, and URLs. This year we are incorporating ethical considerations in the review process. Authors will be allowed extra unlimited space after the main content for a broader impact statement or other discussions of ethics. Please review the ethics policy before submitting. =============================== Important Dates =============================== - Direct paper submission deadline: Friday, 2 December 2022 - Notification of acceptance: Friday, 8 February 2023 - Camera-ready papers due: Wednesday, 1 March 2023 - Main Conference: Wednesday – Friday, 3–5 May 2023 All deadlines are 11.59 pm UTC -12h ("anywhere on Earth"). ======================== Paper Submission Information ========================= --- Topics of Interest --- Of interest are all topics related to theoretical and applied computational linguistics, such as (but not limited to) the topics listed on the main conference website. Submitted systems papers may be of the following types: - Natural Language Processing systems or system components - Application systems using language technology components - Software tools for Computational Linguistics research - Software for analysis, demonstration, or evaluations - Software supporting learning or education - Tools for data visualization and annotation - Development tools Please note: Commercial sales and marketing activities are not appropriate in the system demonstration track at EACL 2023. --- Submission Guidelines --- All submissions should be made electronically via START. Please note that the submission platform is different from the one used by the main conference. Submissions must include the following: - A paper describing the motivation and the technical details of the system, including visual aids (e.g., screenshots, snapshots, or diagrams). We encourage authors to check recent demonstration papers at previous ACL, EACL, EMNLP, and NAACL conferences for examples. - A short video (max. 2 minutes) demonstrating the system. This video will be used to evaluate the paper, but it will not be published unless requested. A screencast with audio narration is a natural choice for demos that can be presented on a screen. Otherwise, a video of a user interacting with the system can be used. The production quality of the video is not of interest. Hence, we encourage the videos to be simply a screencast of the software that is getting demoed, with zero to minimal editing efforts. We recommend that you publish your video on YouTube, or another website and include the link in your paper. If you prefer not to publicly upload a screencast, please submit the video (in MPEG4 format). The video must be included as supplementary material when you submit your paper through START. In addition, we strongly recommend that all demos be made available via one of the following formats: (a) a live demo website, or (b) a website with a downloadable installation package of the demo. We understand though that this might be impossible, e.g., when special hardware is required or when access is otherwise limited. --- Submission Policy --- The demo paper has to be original, written specifically for this conference, and cannot be submitted elsewhere. The paper must also report on a substantial improvement (>30%) if the system that is being described has been reported elsewhere before. Authors submitting more than one demo paper to EACL 2023 must ensure that the submissions do not overlap significantly (>25%) with each other. Submissions of identical or closely related work to multiple tracks at EACL 2023 (Research, SRW, Industry, or Demo) will be rejected by all tracks. --- Reviewing Policy --- Reviewing will be single-blind, and thus authors do not need to conceal their identity. Thus, the demo papers should include the authors’ names and affiliations. Self-references are also allowed. Relevant papers that meet formatting requirements will be assessed on the basis of their relevance to the demo track, contribution, clarity, completeness, and novelty. --- Ethics Policy --- Authors are required to honor the ethical code set out in the ACM Code of Ethics. The consideration of the ethical impact of our research, use of data, and potential applications of our work has always been an important consideration, and as artificial intelligence is becoming more mainstream, these issues are increasingly pertinent. We ask that all authors read the code, and ensure that their work is conformant to this code. We reserve the right to reject papers on ethical grounds, where the authors are judged to have operated counter to the code of ethics, or have inadequately addressed legitimate ethical concerns about their work. Authors will be allowed extra space after the 6th page for a broader impact statement or other discussions of ethics. The ACL demonstration review form will include a section addressing these issues and papers flagged for ethical concerns by reviewers will be further reviewed by an ethics committee. Note that an ethical considerations section is not required, but papers working with sensitive data or on sensitive tasks that do not discuss these issues will not be accepted. Conversely, the mere inclusion of an ethical considerations section does not guarantee acceptance. In addition to acceptance or rejection, papers may receive a conditional acceptance recommendation. Camera-ready versions of papers designated as conditional acceptance will be re-reviewed by the ethics committee to determine whether the concerns have been adequately addressed. Please read the ethics FAQ (shared with the main conference) for more guidance on some problems to look out for and key concerns to consider relative to the code of ethics. ---- Best Demo Award ---- We will present a Best Demo Paper Award. The winner will be chosen based on the contribution and the completeness of the system, as assessed by the reviewers and also based on the live demo at the conference. ============================= Contact Information ============================= If you have questions that are not answered there, please email the program co-chairs at eacl2023demos(a)googlegroups.com. ---- Demo Track Co-chairs ---- - Danilo Croce (University of Rome, Tor Vergata) - Luca Soldaini (Allen Institute for AI)

1 0

2025

2024

2023

2022

Corpora