Center for Language & Speech Processing(CLSP), JHU Courses

Condition-Aware Self-Supervised Learning Representation for Generalized Speech Processing

1 rating

Discover how Condition-Aware Self-Supervised Learning Representation enhances speech processing through dynamic language and speaker embeddings, improving performance across multiple speech-related tasks.

• YouTube
• 26 minutes
• Self-Paced
• Free Video

Multi-scale Analysis and Visualization of Pitch Structures in Music

1 rating

Explore multi-scale analysis techniques for visualizing pitch structures in musical compositions through computational methods and data representation.

• YouTube
• 1 hour 11 minutes
• Self-Paced
• Free Video

Direct Models for Word Alignment and Machine Translation - Arabic to English

1 rating

Explore direct models for word alignment and machine translation, focusing on Arabic-English applications. Learn about supervised training, improved alignments, and recent advancements in the field.

• YouTube
• 1 hour 2 minutes
• Self-Paced
• Free Video

TurkingBench: A Challenge Benchmark for Web Agents - NAACL 2025

Explore TurkingBench, a benchmark testing multi-modal models on web-based tasks using natural HTML pages from crowdsourcing platforms, with a framework for evaluating chatbot responses to web actions.

• YouTube
• 10 minutes
• Self-Paced
• Free Video

Benchmarking Language Model Creativity: A Case Study on Code Generation

Explore a framework for measuring LLM creativity in code generation through denial prompting and the NeoGauge metric, revealing how even GPT-4 falls short of human-like creative problem-solving.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

Scalable Training for Machine Translation Made Successful for the First Time

Discover groundbreaking advances in machine translation training through a violation-fixing perceptron framework that enables scalable processing with 20M+ sparse features, achieving significant BLEU score improvements.

• YouTube
• 1 hour 4 minutes
• Self-Paced
• Free Video

HLTCOE Submission to the VoicePrivacy Attacker Challenge - Improving ASV Performance Against Anonymized Speech

Discover effective methods to improve ASV performance against anonymized speech, including classifier enhancements, alternative distance metrics, and kNN-VC normalization to reduce EER in voice privacy systems.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

What Roles Will Humans Play in the Future of Data Annotation?

Explore the evolving relationship between humans and large language models in data annotation, examining whether traditional crowdsourcing can outperform LLMs and how humans effectively prompt these models without gold-standard references.

• YouTube
• 1 hour 8 minutes
• Self-Paced
• Free Video

Evaluating and Inducing Dialectal Robustness in Large Language Models

Explore how language models perform across dialects and variants, with a focus on DialUp, a method to induce robustness to dialect continua in machine translation models, including for unseen dialects.

• YouTube
• 53 minutes
• Self-Paced
• Free Video

Development and Deployment of Large-Scale AI Systems and Tasks They Still Struggle With

Explore the development and evaluation of large language models, focusing on data generation challenges, multi-model integration strategies, and areas where advanced AI systems still face significant limitations.

• YouTube
• 1 hour 9 minutes
• Self-Paced
• Free Video

Cross-layer Models for Low-Resource Conversational ASR

Discover cross-layer models integrating pronunciation and prosodic variation for conversational ASR, especially effective in low-resource settings and applications in healthcare.

• YouTube
• 1 hour 31 minutes
• Self-Paced
• Free Video

Cross-Layer Models for Low-Resource Conversational ASR

Explore cross-layer models integrating pronunciation and prosodic variation for conversational ASR, especially effective in low-resource settings and healthcare applications.

• YouTube
• 1 hour 31 minutes
• Self-Paced
• Free Video

Differentiable Modeling for Machine Learning

Explore differentiable forward modeling techniques that integrate scientific knowledge with deep learning for sparse data domains and inverse problem solving.

• YouTube
• 1 hour 21 minutes
• Self-Paced
• Free Video

Differentiable Modeling for Machine Learning

Explore differentiable forward modeling in machine learning, combining scientific knowledge with deep learning for sparse data domains like computer graphics and acoustics.

• YouTube
• 1 hour 21 minutes
• Self-Paced
• Free Video

Large Concept Model - Beyond Token-Based LLMs

Discover Large Concept Models that transcend token-based LLMs through multimodal reasoning, hierarchical processing, and zero-shot multilingual capabilities in this Meta AI research presentation.

• YouTube
• 1 hour 22 minutes
• Self-Paced
• Free Video