NLP & LLMs: Mastering Arabic Language Intricacies
summary
Claims that modern NLP and LLM approaches can handle Arabic morphology and mixed Arabic-English text by using a released Annotated Arabic-English Receipt Dataset and a prototype Key Information Extraction and Classification pipeline; shows the end-to-end system (data, tokenization, model fine-tuning, post-processing) and reports qualitative gains on noisy receipts, including error cases where right-to-left and mixed-script tokens break off-the-shelf models. Demonstrates a working bilingual KIE pipeline and practical preprocessing tricks that reduce labeling noise.
AI-written summary · about
related talks
18mLLMs and GenAI: A Guide to Building Your Own RAG Model for Document Interaction
49mA Practical Approach to Using LLMs: Peculiarities and Pitfalls
32mDocuments Meet LLMs: Tales from the Trenches
52mAnatomy of an LLM
19mLLM-Enhanced Multimodal AI: Revolutionizing Audio/Video Interaction
01h 19mBuilding an LLM on your Laptop to...
01h 39mMastering LLMs: From Prompt Engineering to Agentic AI
57mLLMs gone wild