Intent Detection in Arabic

def prune_transformer(model, amount=0.3): for name, module in model.named_modules(): if isinstance(module, torch.nn.Linear): prune.l1_unstructured(module, name='weight', amount=amount) prune.remove(module, 'weight') return model # Efficient Multilingual NMT class LightTranslator(nn.Module): def __init__(self, vocab_size, d_model=512): super().__init__() self.encoder = TransformerEncoder(d_model) self.decoder = TransformerDecoder(d_model) def forward(self, src, tgt): memory = self.encoder(src) return self.decoder(tgt, memory)

Project Overview

Built intent classification models for Arabic conversational and query text at CERIST, a core component for downstream applications such as chatbots, voice assistants, or automated support systems operating in Arabic.

Technical Details

Architectures: Fine-tuned AraBERT, consistent with the sentiment analysis work, compared against classical baselines.
Intent taxonomy: A compact set of domain-specific intents, sized to the target application rather than to a general-purpose assistant.
Data: Short Arabic queries and conversational turns, collected and labeled against that taxonomy.

Complements the local LLM and RAG deployment work by handling structured intent classification ahead of retrieval or generation.

Intent Detection in Arabic

Project Overview

Technical Details

Related Work