TMT: Tri-Modal Translation between Speech, Image, and Text by Processing Different Modalities as Different Languages

TL;DR

This paper introduces TMT, a tri-modal translation model that treats speech, image, and text as different languages, enabling efficient and unified translation across these modalities with improved performance.

Contribution

The paper presents a novel tri-modal translation framework that interprets different modalities as languages, reducing computational costs and enhancing multi-modal translation performance.

Findings

TMT outperforms single-model counterparts on all six modality translation tasks.

Tokenization into discrete tokens enables a unified interface across modalities.

Treating modalities as languages benefits both practicality and accuracy.

Abstract

The capability to jointly process multi-modal information is becoming an essential task. However, the limited number of paired multi-modal data and the large computational requirements in multi-modal learning hinder the development. We propose a novel Tri-Modal Translation (TMT) model that translates between arbitrary modalities spanning speech, image, and text. We introduce a novel viewpoint, where we interpret different modalities as different languages, and treat multi-modal translation as a well-established machine translation problem. To this end, we tokenize speech and image data into discrete tokens, which provide a unified interface across modalities and significantly decrease the computational cost. In the proposed TMT, a multi-modal encoder-decoder conducts the core translation, whereas modality-specific processing is conducted only within the tokenization and detokenization stages. We evaluate the proposed TMT on all six modality translation tasks. TMT outperforms single model counterparts consistently, demonstrating that unifying tasks is beneficial not only for practicality but also for performance.