PaperWave: Listening to Research Papers as Conversational Podcasts Scripted by LLM

TL;DR

PaperWave investigates transforming research papers into conversational podcasts via LLMs to enable mobile, listening-based engagement with niche scholarly content. Through fieldwork, autobiographical design, and a design workshop with 11 participants, the study reveals that podcast-style papers can lower barriers to engagement and shift attention to different aspects of papers, while raising concerns about accuracy and missing visuals. The authors prototype three iterations (manual ChatGPT, CLI automation, and a web app) to deliver PDF-to-audio workflows with configurable language, duration, and playback. Findings emphasize the need to consider listener-environment interaction and audience variation when designing document-to-audio systems, as well as the potential for broader topic exploration through sharing and mobile listening. Limitations include non-generalizability, potential biases, and lack of personalization, pointing to a future research space integrating reading-support and multimodal augmentation.

Abstract

Listening to audio content, such as podcasts and audiobooks, is one way for people to engage with knowledge. Listening affords people more mobility than reading by seeing, thereby broadening their learning opportunities. This study explores the potential applications of large language models (LLMs) to adapt text documents to audio content and addresses the lack of listening-friendly materials for niche content, such as research papers. LLMs can generate scripts of audio content in various styles tailored to specific needs, such as full-content duration or speech types (monologue or dialogue). To explore this potential, we developed PaperWave as a prototype that transforms academic paper PDFs into conversational podcasts. Our two-month investigation, involving 11 participants (including the authors), employed an autobiographical design, a field study, and a design workshop. The findings highlight the importance of considering listener interaction with their environment when designing document-to-audio systems.

Figures (4)

• Figure 1: Recording page. Users can adapt research paper PDFs into conversational podcasts using the interface shown in this figure (this adaptation is referred to as recording in the app). (A) Upload a PDF file to adapt. (B) Enter the title of the episode to record. (C) Set the duration of the episode. (D) Choose the language of the speakers in the podcast episode. (E) Choose the LLM model to generate the script. (F) Advanced options include episode description, keywords, and cover image URL.
• Figure 2: Episodes page shows a list of recorded podcasts. (A) Episodes page on large screens. (B) After the user has input the PDF, the episode will be displayed with a recording status until the generation is complete. Depending on the duration specified by the user, the recording will take about five minutes. (C) When the recording is complete, the player interface will be displayed. Users will use this interface to playback episodes. (D) Episodes page on mobile devices. All pages of PaperWave support responsive design and can be accessed from mobile devices. Users can listen to episodes in various locations while doing everyday tasks, such as chores or traveling, with their mobile devices.
• Figure 3: Channels page shows a list of episodes created by colleagues. Users can select a channel and visit the colleague's episodes page to listen to the episodes recorded by the colleague.
• Figure 4: PaperWave CLI implementation. The boxed texts show the instructions for the LLM for adaptation. Process from the input of the PDF to the audio of the podcast is illustrated.