Sentence Reordering

Introduction

Project for the Deep Learning course at the University of Bologna (A.Y. 2023-2024).

Description

Autoregressive Transformer encoder-decoder model implemented in Keras and trained to reorder a shuffled text sequence.

Results

The submitted model has $\sim 5 \text{M}$ parameters and reaches a score (based on the longest common substring) of $\sim 0.51$.