Audio-Novel

Analysis of using GPT-3 as a co-narrator in live improvisational theatre. Covers methodology, audience/performer feedback, and implications for human-AI creative collaboration.

Analysis of LLMs' ability to maintain consistent fictional worlds, revealing limitations in narrative coherence and state retention for creative writing.

Comprehensive technical documentation and resources about audio-novel technology and applications.

Research on retrieving audio using free-form natural language queries, introducing new benchmarks and baselines for cross-modal audio retrieval.

An analysis of the theoretical framework for conceptualizing differences between reading printed books and listening to audiobooks, emphasizing mobile listening practices.

Analysis of AudioBoost, a system using LLMs to generate synthetic queries from audiobook metadata to improve retrieval and query suggestions in Spotify's cold-start scenario.

Analysis of AudioBoost, a system using Large Language Models to generate synthetic queries for improving audiobook retrievability in Spotify's search engine during cold-start scenarios.

Analysis of Audiobook-CC, a novel speech synthesis framework for generating coherent, emotionally expressive multicast audiobooks with fine-grained control and long-context modeling.

Analysis of Audiobook-CC, a novel TTS framework for generating coherent, emotionally expressive, and contextually consistent multicast audiobooks with fine-grained control.

A critical analysis of the book review on 'Digital Audiobooks' exploring mediatization theory, post-phenomenology, and the evolving landscape of audio-based literature consumption.

Analysis of end-to-end speech-to-text translation models on an augmented audiobook corpus, exploring training scenarios and model efficiency.

An in-depth analysis of film music through the lens of narratology, exploring how music functions across different narrative levels in cinema.

Spotify's innovative 2T-HGNN system using Graph Neural Networks for audiobook recommendations, achieving +46% start rate and +23% streaming rate improvements.

A novel approach for weakly-supervised temporal grounding of procedural steps in instructional videos using multi-modal alignment of frames, narrations, and step descriptions from wikiHow.

Analysis of J-MAC corpus construction methodology, technical contributions, evaluation results, and future directions for expressive audiobook speech synthesis.

A novel meta-learning approach (MAMLCon) that mitigates catastrophic forgetting in continual few-shot learning for spoken word classification, outperforming existing methods like OML.

Analysis and framework for integrating Mobile Audiobooks (MABs) to develop listening comprehension skills in EFL college students, covering advantages, selection, implementation, and assessment.

Analysis and framework for integrating mobile audiobooks to develop listening comprehension skills in EFL college students, covering advantages, selection, implementation, and assessment.

Analysis of Movie101v2, a large-scale bilingual dataset for movie narration, including its three-stage task roadmap, baseline evaluations, and future research directions.

An analysis of research investigating how background music with and without lyrics impacts spoken-word recognition, with implications for social settings and future work.

An interactive authoring tool that simplifies the creation of animated data videos by linking narration to chart animations and providing structure-aware animation effects.

A research paper introducing the task of automatic narration generation for videos, presenting a new dataset from Peppa Pig, and proposing models for timing and content generation.

Analysis of a computational model exploring phonetic perception's role in non-native word processing, challenging traditional phonological explanations.

A two-stage framework for embedding spoken words with both phonetic and semantic information, enabling advanced spoken document retrieval beyond simple term matching.

Research on predicting prosody attributes (pitch, volume, rate) from narrative text using NLP and language models, improving TTS for audiobook generation.

Research on computational identification of unreliable narrators using LLMs, featuring the TUN A dataset and classification of intra-narrational, inter-narrational, and inter-textual unreliability.

A research paper exploring how to use noisy audio narration as weak supervision to train action detection models, reducing annotation costs while leveraging multimodal video features.

A zero-shot approach for generating expressive audiobooks using multimodal speaker personas and LLM-based script instructions, eliminating costly training and manual annotation.

A technical analysis of MultiActor-Audiobook, a novel zero-shot system for generating expressive audiobooks using multimodal speaker personas and LLM-based script instructions.