Welcome to the repository for the MAOSTE, a novel task in the field of multimodal aspect-based sentiment analysis. This research work introduces the task of extracting triples of aspects, objects, and sentiments from multimodal data sources.

• maoste/: This directory contains the codebase for the MAOSTE project. Inside, you'll find all the necessary scripts and modules required to set up, train, and evaluate the model.

• model_train_result/: In this folder, you will find the trained models along with the prediction results. It includes the weights of the neural networks, the configuration files, and the evaluation metrics calculated during the training process.

• original_dataset/: This directory hosts the original dataset used for training and evaluating the MAOSTE model. It includes raw data, preprocessed inputs, and the ground truth annotations necessary for training and testing the model.

We are committed to supporting and expanding the MAOSTE project. In line with this commitment, we plan to progressively release the remaining experimental code, data, and other relevant materials. These updates will further enhance the utility and applicability of the MAOSTE project for researchers and practitioners in the field.

Stay tuned for these additions, which will provide more comprehensive insights into our methodologies and facilitate deeper engagement with the tasks of multimodal aspect-based sentiment analysis.

We encourage you to regularly check back for these updates or watch/star this repository to stay informed about new releases.