David Ifeoluwa Adelani 🇳🇬
@davlanade
Followers
3K
Following
12K
Media
163
Statuses
4K
Assistant Professor @mcgillu, Core Academic Member @Mila_Quebec, Canada CIFAR AI Chair @CIFAR_News | interested in multilingual NLP | Disciple of Jesus
Saarbrücken, Germany
Joined September 2017
I'm happy to share the electronic copy of my PhD thesis ."Natural language processing for African languages" . Happy reading especially if you are new to AfricaNLP. A big thank you to my supervisor @dklakow & the examination committee members:
Thank you @NaijaNLP . Hallelujah! I'm happy to share that I defended my PhD today on "Natural language processing for African languages". The journey of 8 years is finally over. A big thank you to my supervisor @dklakow and reviewers: Prof. Alex Fraser & Prof. Benoît Sagot.
35
80
352
RT @aremuadeolajr: Last week, I gave a talk on The State of African NLP: Africa has 2000+ languages, yet most remain low-resource in AI. On….
0
7
0
RT @jurafsky: Now that school is starting for lots of folks, it's time for a new release of Speech and Language Processing! Jim and I adde….
0
68
0
RT @vukosi: We gathered in Kigali 🌍 for the 2025 Deep Learning Indaba. Nearly 3,000 applied, 1,294 joined in person, 400 virtually. 6 days….
0
43
0
RT @ziling_cheng: Our paper on reasoning × interpretability × evaluation has been accepted to EMNLP main! . Excited because this marks the….
0
13
0
5 papers accepted at #EMNLP2025: pp below. - SSA-COMET @SenyuLiReal. - AfriDOC-MT &.African NLP survey @alabi_jesujoba. - TransWebLLM . - AfroXLMR-Social
arxiv.org
Language models built from various sources are the foundation of today's NLP progress. However, for many low-resource languages, the diversity of domains is often limited -- more biased to a...
8
9
118
RT @Yoshua_Bengio: It was a pleasure to meet Prime Minister @MarkJCarney and Minister @EvanLSolomon today as part of an official visit at @….
0
18
0
RT @ArxivSound: ``mSTEB: Massively Multilingual Evaluation of LLMs on Speech and Text Tasks,'' Luel Hagos Beyene, Vivek Verma, Min Ma, Jesu….
arxiv.org
Large Language models (LLMs) have demonstrated impressive performance on a wide range of tasks, including in multimodal settings such as speech. However, their evaluation is often limited to...
0
3
0
0
0
2
We performed error analysis on the LID task: confusion matrix reveal that many of the LID errors are with languages from the same region. For example, Xhosa vs Zulu. Occitan & Asturian have very low LID accuracy in the Europe region, followed by luxembourgish, Irish & Bosnia.
1
0
2
Audio LLMs perform the worst for African and Americas/Oceania language across all tasks, however Gemini 2.0 Flash perform better on low-resource languages than GPT-4o Audio. Qwen 2 Audio only perform well on few Asian and European languages.
1
0
1
We compared the performance of 3 audio LLMs: Qwen 2 Audio, Gemini 2 Flash and GPT-4o Audio on various tasks: LID, ASR, S2TT, topic classification and reading comprehension. [speech vs text] There is wider gap between open & proprietary models on speech than on text-based tasks
1
0
1
I'm excited to share that our paper ."mSTEB: Massively Multilingual Evaluation of LLMs on Speech and Text Tasks" got accepted at #ASRU2025. We provide a large scale eval. for speech tasks on over 100 languages. leaderboard:
1
14
81
Fleurs-SLU accepted at #COLM2025 . See you in Montréal. See the pre-print
arxiv.org
Spoken language understanding (SLU) is indispensable for half of all living languages that lack a formal writing system. Unlike for high-resource languages, for these languages, we cannot offload...
Fleurs-SLU benchmark is out 🎉, this is addressing one of the major limitations of speech representation evaluation. We release 2 benchmark datasets based on:.- SIB-200 (topic classification) & .- Belebele (reading comprehension). It was great working with @fdschmidt on this.
0
2
20
RT @jxmnop: OpenAI hasn’t open-sourced a base model since GPT-2 in 2019. they recently released GPT-OSS, which is reasoning-only. or is….
0
471
0
RT @gneubig: Multimodal LLMs lack cultural knowledge, especially for less well-represented cultures and languages. To fix this, we created….
0
13
0
RT @JessicaOjo__: With these new GPT drops, I have to see how they handle low-resource languages - especially African languages. Running A….
0
3
0
RT @maksym_andr: 🚨 Incredibly excited to share that I'm starting my research group focusing on AI safety and alignment at the ELLIS Institu….
0
88
0
RT @Eric_Wallace_: Today we release gpt-oss-120b and gpt-oss-20b—two open-weight LLMs that deliver strong performance and agentic tool use.….
0
381
0
RT @gneubig: Summary of GPT-OSS architectural innovations:. 1. sliding window attention (ref: .2. mixture of expert….
arxiv.org
Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly,...
0
359
0