Multimodal Sentiment Analysis for 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

Extended sentiment analysis beyond text alone by incorporating additional modalities (e.g. audio tone, visual cues) for Arabic content, addressing cases where sentiment expressed in speech or video is not fully captured by transcribed text, such as sarcasm or dialectal expressiveness carried in tone.

Technical Details

Modalities: Text combined with audio and visual features.
Fusion approach: Learned fusion over modality-specific encoders, rather than plain feature concatenation.
Data: Arabic speech and video content with sentiment labels. Arabic multimodal corpora are scarce enough that assembling usable data was part of the work itself.

Builds directly on Arabic Sentiment Analysis with Transformers, extending the text-only approach to additional modalities.

Multimodal Sentiment Analysis for Arabic

Project Overview

Technical Details

Related Work