Improving the Cross-Lingual Generalisation in Visual Question Answering

• DOI: 10.1609/aaai.v37i11.26574
• Publication Date: 2023-06-27T18:08:39Z
• Authors: Farhad Nooralahzadeh; Rico Sennrich
• Keywords: FOS: Computer and information sciences; Computer Science - Computation and Language; 10105 Institute of Computational Linguistics; 0202 electrical engineering, electronic engineering, information engineering; 410 Linguistics; 02 engineering and technology; 000 Computer science, knowledge & systems; Computation and Language (cs.CL); 11551 Zurich Center for Linguistics

While several benefits were realized for multilingual vision-language pretrained models, recent benchmarks across various tasks and languages showed poor cross-lingual generalisation when multilingually pre-trained vision-language models are applied to non-English data, with a large gap between (supervised) English performance and (zero-shot) cross-lingual transfer. In this work, we explore the poor performance of these models on a zero-shot cross-lingual visual question answering (VQA) task, where models are fine-tuned on English visual-question data and evaluated on 7 typologically diverse languages. We improve cross-lingual transfer with three strategies: (1) we introduce a linguistic prior objective to augment the cross-entropy loss with a similarity-based loss to guide the model during training, (2) we learn a task-specific subnetwork that improves cross-lingual generalisation and reduces variance without model modification, (3) we augment training examples using synthetic code-mixing to promote alignment of embeddings between source and target languages. Our experiments on xGQA using the pretrained multilingual multimodal transformers UC2 and M3P demonstrates the consistent effectiveness of the proposed fine-tuning strategy for 7 languages, outperforming existing transfer methods with sparse models.