PA-TMM

This is a PyTorch implementation of the paper "Responding to News Sensitively in Stock Attention Networks via Prompt-Adaptive Trimodal Model".

Requirements

News headlines are collected from Benzinga.

Run MPA.ipynb.
This script would build and pre-train a PA-TMM model with the movement prompt adaptation strategy.
Run PA-TMM.ipynb.
This script would load the pre-trained parameters to the PA-TMM model, and then fine-tune it.