Pré-Publication, Document De Travail Année : 2024

Visual Question Generation on VQA Dataset

Xueting Pan
  • Fonction : Auteur

Résumé

Question Generation in reading comprehension has garnered significant interest recently. Traditionally, most research in this area has focused on generating questions from textual information related to the reading passage using sequence-to-sequence models. In this work, we explored an alternative approach by investigating the generation of questions solely from images that may accompany a reading passage. This novel approach involves leveraging a deep learning model that combines VGG (Visual Geometry Group) and LSTM (Long Short-Term Memory) networks in a serial architecture. By training the model on a Visual Question Answering (VQA) dataset, it has been demonstrated that the system can effectively generate insightful and contextually relevant questions from images sourced from the internet. This research highlights the potential for integrating visual content into the question generation process, offering new avenues for enhancing reading comprehension tasks.

Fichier principal
Vignette du fichier
Visual_Question_Generation_on_VQA_Dataset (1).pdf (3.14 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-04667513 , version 1 (04-08-2024)

Licence

Identifiants

  • HAL Id : hal-04667513 , version 1

Citer

Ziqian Luo, Xueting Pan. Visual Question Generation on VQA Dataset. 2024. ⟨hal-04667513⟩
129 Consultations
287 Téléchargements

Partager

  • More