Muhammad AbdulMageed @mageed

Our work "Alexandria: A Multi-Domain Dialectal Arabic Machine Translation Dataset for Culturally Inclusive and Linguistically Diverse LLMs " will be presented by @mageed at #ACL2026 Main. 📍 Coronado: Session 17: Poster Session G 🗓 Tues., July 7: 11:00 AM–12:30 PM @aclmeeting

@ubcscience @UBC_Arts @UBCLinguistics @UBC_NLP @UBC_CS @UBCiSchool @UBCLangScis

🌍 We're heading to San Diego to present five papers at ACL 2026. This year’s work from our group and collaborators spans dialectal Arabic machine translation, speech technology, machine translation evaluation, context-aware ASR, and efficient reasoning distillation. I am especially grateful to the many students, collaborators, and community contributors whose work made these papers possible. 1. Alexandria: A Multi-Domain Dialectal Arabic Machine Translation Dataset for Culturally Inclusive and Linguistically Diverse LLMs Authors: Abdellah EL Mekki (Abdellah EL Mekki, Ph.D.), Samar M. Magdy (Samar M. Magdy), Houdaifa Atou (Houdaifa Atou), Ruwa AbuHweidi (Ruwa AbuHweidi), Baraah Qawasmeh, OMER NACAR, Thikra Al-hibiri, Razan Saadie, Hamzah Alsayadi, Dr. Eng. Nadia GHEZAIEL HAMMOUDA, Alshima Mohammed Alkhazimi, Aya Hamod, Al-Yas Yaqoob Al-Ghafri, Wesam Ali El-Sayed, Asila Ismail al Sharji, Mohamad Ballout, Anas Belfathi, Karim Ghaddar, Serry Sibaee, Alaa Aoun, Aeej Mohammed Aseri, Lina Abureesh, Ahlam Bashiti, Majdal Yousef, Abdulaziz Hafiz, Yehdih Mohamed, Emira Hamedtou, Brakehe Emehah, Rahaf Alhamouri, Youssef Nafea, Aya El Aatar, Walid Al-DHabyani, Emhemed Hamed, Sara Shatnawi, Fakhraddin Alwajih, Khalid Elkhidir, Ashwag Alasmari, Abdurrahman Gerrio, Omar Said Alshahri, AbdelRahim Elmadany, Ismail Berrada, Amir Azad Adli Al-kathiri, Fadi Zaraket, Mustafa Jarrar, Yahya Mohamed Elhadj, Hassan Alhuzali, Muhammad Abdul-Mageed URL: lnkd.in/gfhiKBjR 2. LQM: Linguistically Motivated Multidimensional Quality Metrics for Machine Translation Authors: Samar M. Magdy, Fakhraddin Alwajih, Abdellah El Mekki, Wesam El Sayed, Muhammad Abdul-Mageed URL: lnkd.in/g6rz7gui 3. Zero-Shot Context-Aware ASR for Diverse Arabic Varieties Authors: Bashar Talafha, Amin Abu Alhassan, Muhammad Abdul-Mageed URL: lnkd.in/gKGyGQgn 4. Arab Voices: Mapping Standard and Dialectal Arabic Speech Technology Authors: Peter Sullivan, AbdelRahim Elmadany, Alcides Alcoba Inciarte, Muhammad Abdul-Mageed URL: lnkd.in/gMzYhT8x 5. Distilling the Essence: Efficient Reasoning Distillation via Sequence Truncation Authors: Wei-Rui Chen, Vignesh Kothapalli, Ata Fatahibaarzi, Hejian Sang, Shao Tang, Qingquan Song, Zhipeng Wang, PhD, Muhammad Abdul-Mageed URL: lnkd.in/ghYp5S_u Congratulations to all co-authors and collaborators. Looking forward to discussing these works with the community at ACL 2026! #ACL2026 #NLProc #ArabicNLP #SpeechTechnology #MachineTranslation #LLMs #ResponsibleAI #MultilingualNLP

keyboard_arrow_left Previous keyboard_arrow_right Next

🙌 Machine translation is an important technology that is becoming ever more pervasive in business and daily life. It is widely known that machine translation errors can be embarrassing, if not catastrophic at times. When a model makes an error, it is not just about using a word or phrase that it should not have used. Rather, the error may reflect a broader blindness to language variation, linguistic structures, cultural norms, situational nuances, and more. For a long time, it has been difficult to characterize machine translation errors methodically and in ways that adhere to linguistic theory. Not anymore. Today, we introduce LQM, a linguistically motivated framework for characterizing machine translation quality. Along with the framework, we construct a bidirectional parallel corpus of machine translation errors, covering seven major Arabic dialects and labeled according to our taxonomy. The corpus is derived from conversational, culturally rich content. We use the corpus to evaluate the machine translation ability of six different LLMs against expert span-level human annotations, producing 6,113 labeled error spans across 3,495 unique erroneous sentences, along with severity-weighted quality scores. Our accompanying paper will appear in ACL 2026 Findings. Read the paper: “LQM: Linguistically Motivated Multidimensional Quality Metrics for Machine Translation” Link: aclanthology.org/2026.findings-… Work by Samar M. Magdy, Fakhraddin AlWajih, Abdellah EL Mekki, and Wesam Ali El-Sayed. #MachineTranslation #MachineTranslationErrors #MTEvaluation #ArabicNLP #ArabicDialects #LLMs #MultilingualNLP #ComputationalLinguistics #NLProc

keyboard_arrow_left Previous keyboard_arrow_right Next

Canada 🇨🇦 is a land of proud creators. Let’s keep building, together. Happy #CanadaDay! 🍁

ISED @ISED_CA

a day ago

Happy #CanadaDay!🍁 Today we celebrate the innovators, researchers, entrepreneurs and communities shaping Canada’s future with bold ideas and real impact. canada.ca/made-in-canada

0 9 19 731 1

Looking forward to receiving submissions to ArabicNLP2026.

ArabicNLP2026 @_ArabicNLP

2 weeks ago

📢 مؤتمر ArabicNLP 2026 — تذكير مهم | آخر موعد لتقديم الملخصات غدًا 📢 نذكّر الباحثين والمهتمين بمجالات معالجة اللغة العربية الطبيعية بأن آخر موعد لتقديم الملخصات لمؤتمر ArabicNLP 2026 هو غدًا: * آخر موعد لتقديم الملخصات: 18 يونيو 2026 * آخر موعد لتقديم الأوراق الكاملة: 25 يونيو

0 3 8 641 1

Yesterday was an exciting milestone with the release of Canada’s AI Strategy. A strong step toward advancing innovation and responsible AI adoption in healthcare.

University Health Network @UHN

4 weeks ago

Proud to host PM @MarkJCarney, Diana Fox Carney and @EvanLSolomon. UHN is excited to partner in shaping the future of AI-enabled health care .

5 21 65 7K 4

@ubcscience @UBC_CS @UBC_Arts @UBCLangScis @UBCLinguistics @UBCiSchool @UBCNews

Canada’s AI strategy is at a pivotal crossroads. I thoroughly enjoyed sitting down with Keri Adams on @CTVNews @YourMorning today to unpack what this means for our future. Navigating the rapid evolution of generative AI requires clear, actionable frameworks, and having these public conversations is vital. Thank you to the CTV team for a fantastic and timely discussion! #ArtificialIntelligence #CanadaAIStrategy #GenerativeAI #MachineLearning #UBC ctvnews.ca/vancouver/vide…

🔗 nadi.dlnlp.ai/2026/

We’re organizing the NADI shared task for the seventh year in a row. NADI 2026 will focus on speech technology, with a wide range of tasks. Tasks include real-world automatic speech recognition, speech dialect identification, text-to-speech, and speech translation. Join us!

Link: mumageed.medium.com/building-with-…

I spent the morning today helping my students on a new work on African speech and language technology. We talk about a lot of datasets and tools, and present a lot of numbers. But I didn’t see language names. I saw people. With each new language we add, I saw the people we can help. I thought about the patient whose life we can help save by making health information more accessible. The farmer whose crop we can help protect through timely advice. The teacher we can help reach more students. The parent we can help support, the child we can help teach, and the community worker we can help empower. Here’s a post I wrote several months ago. We have done a lot of work since then, but it delivers my own passion for this social program.

📢 مؤتمر ArabicNLP 2026 - الدعوة الثالثة لتقديم الأوراق البحثية 📢 سيعقد المؤتمر الرابع لمعالجة اللغة العربية الطبيعية (ArabicNLP 2026) بالتزامن مع مؤتمر EMNLP 2026 في بودابست، المجر (أكتوبر 2026). - آخر موعد لتقديم الملخصات: ١٨ يونيو ٢٠٢٦ - آخر موعد لتقديم الأبحاث الكاملة: ٢٥ يونيو ٢٠٢٦ للاطلاع على المزيد من التفاصيل حول تعليمات التقديم وغيرها، تفضل بزيارة الموقع الإلكتروني:arabicnlp2026.sigarab.org/call-for-papers 📢 ArabicNLP 2026 - Third Call for Papers 📢 𝐓𝐡𝐞 𝐅𝐨𝐮𝐫𝐭𝐡 𝐀𝐫𝐚𝐛𝐢𝐜 𝐍𝐚𝐭𝐮𝐫𝐚𝐥 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐏𝐫𝐨𝐜𝐞𝐬𝐬𝐢𝐧𝐠 𝐂𝐨𝐧𝐟𝐞𝐫𝐞𝐧𝐜𝐞 (𝐀𝐫𝐚𝐛𝐢𝐜𝐍𝐋𝐏 𝟐𝟎𝟐𝟔) 𝐰𝐢𝐥𝐥 𝐛𝐞 𝐜𝐨-𝐥𝐨𝐜𝐚𝐭𝐞𝐝 𝐰𝐢𝐭𝐡 𝐄𝐌𝐍𝐋𝐏 𝟐𝟎𝟐𝟔 𝐢𝐧 𝐁𝐮𝐝𝐚𝐩𝐞𝐬𝐭, 𝐇𝐮𝐧𝐠𝐚𝐫𝐲 (𝐎𝐜𝐭𝐨𝐛𝐞𝐫 𝟐𝟎𝟐𝟔) - Abstract Deadline: June 18, 2026 - Full Paper Submission: June 25th, 2026 Visit the website for more details on submission instructions and more: https:arabicnlp2026.sigarab.org/call-for-papers #ArabicNLP2026⁩ #ArabicNLP⁩

Minister Joly announces funding for Talent Innovation Canada, a new organization that will connect top graduate talent with Canadian companies working to solve R&D challenges. Learn more: canada.ca/en/innovation-…

@UBCNews @UBC_Arts @UBCLinguistics @UBCiSchool @UBC_CS @UBC_NLP

𝗧𝗵𝗲 𝗻𝗲𝘅𝘁 𝗳𝗿𝗼𝗻𝘁𝗶𝗲𝗿 𝗶𝗻 𝗔𝗜 𝗶𝘀 𝗻𝗼𝘁 𝗼𝗻𝗹𝘆 𝘁𝗲𝗮𝗰𝗵𝗶𝗻𝗴 𝗺𝗮𝗰𝗵𝗶𝗻𝗲𝘀 𝘁𝗼 𝘀𝗽𝗲𝗮𝗸 𝗺𝗼𝗿𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀. It is teaching them to stop flattening people into language names. “Arabic.” “English.” “Spanish.” ... These labels are convenient for datasets. But humans do not speak labels. We speak from cities, villages, families, histories, professions, memories, migrations, jokes, wounds, rituals, and relationships. A grandmother does not speak like a government form. A patient does not describe pain like a Wikipedia article. A farmer, a nurse, a student, a shopkeeper, a traveler, and a child do not all speak the same “standard” version of a language. And yet, much of modern AI still behaves as if language were uniform, and separable from the people who use it. This is one of the great unsolved problems in multilingual AI. Not only translation. 𝗩𝗮𝗿𝗶𝗮𝘁𝗶𝗼𝗻. 𝗦𝗼𝗰𝗶𝗮𝗹 𝗺𝗲𝗮𝗻𝗶𝗻𝗴. 𝗗𝗶𝗮𝗹𝗲𝗰𝘁. 𝗥𝗲𝗴𝗶𝘀𝘁𝗲𝗿. 𝗖𝘂𝗹𝘁𝘂𝗿𝗲. 𝗖𝗼𝗻𝘁𝗲𝘅𝘁. 𝗪𝗵𝗼 𝗴𝗲𝘁𝘀 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗼𝗼𝗱 𝗯𝘆 𝗺𝗮𝗰𝗵𝗶𝗻𝗲𝘀, 𝗮𝗻𝗱 𝘄𝗵𝗼 𝗴𝗲𝘁𝘀 𝗻𝗼𝗿𝗺𝗮𝗹𝗶𝘇𝗲𝗱 𝗮𝘄𝗮𝘆. This is the global problem behind our new project: 𝗔𝗹𝗲𝘅𝗮𝗻𝗱𝗿𝗶𝗮. The name is intentional. Alexandria was never just a city. It was one of humanity’s boldest symbols of knowledge moving across languages, cultures, and worlds. A place where translation was not clerical work. It was civilization-building. Today, we borrow that name for a new question at the heart of AI: 𝗖𝗮𝗻 𝗺𝗮𝗰𝗵𝗶𝗻𝗲𝘀 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝗔𝗿𝗮𝗯𝗶𝗰 𝗻𝗼𝘁 𝗮𝘀 𝗼𝗻𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲, 𝗯𝘂𝘁 𝗮𝘀 𝗮 𝗹𝗶𝘃𝗶𝗻𝗴 𝘄𝗼𝗿𝗹𝗱 𝗼𝗳 𝘃𝗼𝗶𝗰𝗲𝘀? For too long, Arabic has been flattened in technology. One label: “Arabic.” One assumed standard. One imagined speaker. One simplified linguistic reality. But Arabic is not one voice. Arabic is Cairo and Casablanca. Amman and Muscat. Khartoum and Tunis. Beirut and Riyadh. Sana’a, Tripoli, Nouakchott, Gaza, Damascus, Algiers, Doha, Manama, and beyond. It is dialect. It is register. It is gendered interaction. It is code-switching. It is place, intimacy, politeness, humor, memory, identity, and everyday life. So I am thrilled to share that 𝗔𝗹𝗲𝘅𝗮𝗻𝗱𝗿𝗶𝗮 has been accepted to 𝗔𝗖𝗟 𝟮𝟬𝟮𝟲 𝗠𝗮𝗶𝗻. 𝗔𝗹𝗲𝘅𝗮𝗻𝗱𝗿𝗶𝗮: 𝗔 𝗠𝘂𝗹𝘁𝗶-𝗗𝗼𝗺𝗮𝗶𝗻 𝗗𝗶𝗮𝗹𝗲𝗰𝘁𝗮𝗹 𝗔𝗿𝗮𝗯𝗶𝗰 𝗠𝗮𝗰𝗵𝗶𝗻𝗲 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗶𝗼𝗻 𝗗𝗮𝘁𝗮𝘀𝗲𝘁 𝗳𝗼𝗿 𝗖𝘂𝗹𝘁𝘂𝗿𝗮𝗹𝗹𝘆 𝗜𝗻𝗰𝗹𝘂𝘀𝗶𝘃𝗲 𝗮𝗻𝗱 𝗟𝗶𝗻𝗴𝘂𝗶𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆 𝗗𝗶𝘃𝗲𝗿𝘀𝗲 𝗟𝗟𝗠𝘀 Alexandria is our largest community-built Arabic machine translation project so far: a large-scale, human-translated, multi-domain dataset designed to help AI systems translate Arabic as people actually use it. It brings together: 📷 𝟭𝟯 𝗔𝗿𝗮𝗯 𝗰𝗼𝘂𝗻𝘁𝗿𝗶𝗲𝘀 📷 𝟯𝟰𝗞+ 𝗺𝘂𝗹𝘁𝗶-𝘁𝘂𝗿𝗻 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝘀 📷 𝟭𝟬𝟳𝗞+ 𝗘𝗻𝗴𝗹𝗶𝘀𝗵 📷 𝗱𝗶𝗮𝗹𝗲𝗰𝘁𝗮𝗹 𝗔𝗿𝗮𝗯𝗶𝗰 𝗽𝗮𝗿𝗮𝗹𝗹𝗲𝗹 𝘁𝘂𝗿𝗻𝘀 📷 𝟭𝟭 𝘀𝗼𝗰𝗶𝗮𝗹-𝗶𝗺𝗽𝗮𝗰𝘁 𝗱𝗼𝗺𝗮𝗶𝗻𝘀, including healthcare, education, agriculture, legal/financial services, tourism, logistics, and workplace communication 📷 𝗳𝗶𝗻𝗲-𝗴𝗿𝗮𝗶𝗻𝗲𝗱 𝗰𝗶𝘁𝘆-𝗹𝗲𝘃𝗲𝗹 𝗱𝗶𝗮𝗹𝗲𝗰𝘁 𝗺𝗲𝘁𝗮𝗱𝗮𝘁𝗮 📷 𝘀𝗽𝗲𝗮𝗸𝗲𝗿–𝗮𝗱𝗱𝗿𝗲𝘀𝘀𝗲𝗲 𝗴𝗲𝗻𝗱𝗲𝗿 𝗰𝗼𝗻𝗳𝗶𝗴𝘂𝗿𝗮𝘁𝗶𝗼𝗻𝘀 📷 𝗺𝘂𝗹𝘁𝗶-𝘁𝘂𝗿𝗻 𝗰𝗼𝗻𝘃𝗲𝗿𝘀𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗰𝗼𝗻𝘁𝗲𝘅𝘁 📷 𝗻𝗮𝘁𝘂𝗿𝗮𝗹𝗹𝘆 𝗼𝗰𝗰𝘂𝗿𝗿𝗶𝗻𝗴 𝗰𝗼𝗱𝗲-𝘀𝘄𝗶𝘁𝗰𝗵𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗘𝗻𝗴𝗹𝗶𝘀𝗵, 𝗙𝗿𝗲𝗻𝗰𝗵, 𝗮𝗻𝗱 𝗠𝗦𝗔 📷 𝗲𝘃𝗮𝗹𝘂𝗮𝘁𝗶𝗼𝗻 𝗰𝗼𝗱𝗲 𝗮𝗻𝗱 𝗱𝗮𝘁𝗮 𝗰𝗿𝗲𝗮𝘁𝗶𝗼𝗻 𝗴𝘂𝗶𝗱𝗲𝗹𝗶𝗻𝗲𝘀 to support reproducibility and future community work But Alexandria is not only a dataset. It is a refusal. A refusal to treat dialects as noise. A refusal to treat “standard” language as the only legitimate language. A refusal to build AI systems that understand the center and mishear the margins. Dialects are not broken versions of languages. They are knowledge systems. They carry social meaning. They carry identity. They carry memory. They carry belonging. They carry the subtle difference between being translated and being understood. When AI erases dialects, it does not merely make a linguistic mistake. It erases context. It erases communities. It erases the conditions under which language becomes human. Our findings show a clear gap: current LLMs can often preserve meaning reasonably well, but still struggle to generate fluent, locally authentic dialectal Arabic, especially across fine-grained sub-dialects, domains, and social contexts. That gap matters. It matters when a patient describes symptoms in the words that feel natural to them. It matters when a farmer asks for advice. It matters when a student learns. It matters when a traveler seeks help. It matters when public services are supposed to reach everyone, not only those who speak the “right” variety. Alexandria is our attempt to say something plainly: 𝗧𝗵𝗲 𝗳𝘂𝘁𝘂𝗿𝗲 𝗼𝗳 𝗺𝘂𝗹𝘁𝗶𝗹𝗶𝗻𝗴𝘂𝗮𝗹 𝗔𝗜 𝗰𝗮𝗻𝗻𝗼𝘁 𝗯𝗲 𝗯𝘂𝗶𝗹𝘁 𝗯𝘆 𝗮𝘃𝗲𝗿𝗮𝗴𝗶𝗻𝗴 𝗵𝘂𝗺𝗮𝗻𝗶𝘁𝘆 𝗶𝗻𝘁𝗼 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝗳𝗼𝗿𝗺𝘀. We need systems that can handle variation, not erase it. Systems that can model culture, not smooth it away. Systems that understand that linguistic diversity is not a complication to be removed. It is the world. This is true for Arabic. It is also true globally, for languages with dialects, accents, registers, contact varieties, mixed codes, regional forms, Indigenous varieties, diasporic forms, youth speech, professional speech, and deeply situated cultural meanings. In the old Alexandria, knowledge moved across languages. With this new Alexandria, we hope to help AI move across dialects, communities, and lived realities. This project was only possible because of a remarkable community of collaborators across the Arab world. I am deeply grateful to everyone who contributed to building, translating, validating, evaluating, and shaping this resource. Paper: arxiv.org/abs/2601.13099 Project website: alexandria.dlnlp.ai Dataset: huggingface.co/datasets/UBC-N… Code and guidelines: github.com/UBC-NLP/Alexan… 𝗔𝗜 𝘀𝗵𝗼𝘂𝗹𝗱 𝗻𝗼𝘁 𝗼𝗻𝗹𝘆 𝗹𝗲𝗮𝗿𝗻 𝗺𝗼𝗿𝗲 𝗹𝗮𝗻𝗴𝘂𝗮𝗴𝗲𝘀. 𝗜𝘁 𝘀𝗵𝗼𝘂𝗹𝗱 𝗹𝗲𝗮𝗿𝗻 𝘁𝗼 𝗵𝗲𝗮𝗿 𝗽𝗲𝗼𝗽𝗹𝗲.

keyboard_arrow_left Previous keyboard_arrow_right Next

Great news for the Arabic NLP community! 🎉 📢 Call for Shared Task Proposals is now open! الإعلان الأول للدعوة لتقديم (Shared Tasks) ضمن المؤتمر الرابع للمعالجة الآلية للغة العربية ArabicNLP 2026، والذي سيُعقد بالتزامن مع EMNLP 2026 في بودابست، هنغاريا، خلال الفترة 24–29 أكتوبر 2026. 🗓️ آخر موعد لتقديم Shared Tasks: 25 أبريل 2026 ✅ إشعارات القبول: 2 مايو 2026 🔗 رابط التقديم: forms.gle/gwUJcDyKcQnvNs… Great news for the Arabic NLP community! 🎉 📢 Call for Shared Task Proposals is now open! The first call for Shared Task Proposals for ArabicNLP 2026 is now out. The Fourth Arabic Natural Language Processing Conference will be co-located with EMNLP 2026 in Budapest, Hungary, on October 24–29, 2026. 🗓️ Shared Task proposal submission deadline: April 25, 2026 ✅ Notification of acceptance: May 2, 2026 🔗 Submission link: forms.gle/gwUJcDyKcQnvNs… #ArabicNLP2026 #ArabicNLP #SharedTasks #ArabicNLPSharedTasks #EMNLP2026