Paper Discussion: SPIRIT LM: Interleaved Spoken and Written Language Model

Join us for an engaging session where we'll discuss SPIRIT LM, an innovative multimodal language model designed to blend speech and text seamlessly. Presented by the team at Meta AI, SPIRIT LM pushes the boundaries of language understanding by integrating spoken and written data into a unified model. We will explore its architecture, applications in speech recognition, text generation, and expressivity modelling, and see how it tackles cross-modal tasks such as automatic speech recognition (ASR), text-to-speech (TTS), and speech classification.

Papar: https://arxiv.org/pdf/2402.05755