Determining the approach, scale and feasibility for developing the Tape Archive Analysis Toolkit (TAAT)

This project is to determine the scale and feasibility for developing a bespoke toolset (Tape Archive Analysis Toolkit - TAAT) that will analyse large audio datasets derived from analogue sources to aid researchers in determining new musicological insights and archives in effectively cataloguing and deriving meaning from their digital collections. The Project Lead has worked on two previous AHRC funded research projects which produced 400+ hours of digitised audio files from reel*to-reel tape archives of historical electronic music composers in England. These materials cover final versions of compositions as well as working materials, intermediate materials, intermediate mixes, draft recordings, and pre-existing works from other composers or radio broadcasts. The potential network of relationships posed by this collection is vast and potentially of great value to musicologists, archives, and libraries in developing new insights and communicating musical histories to the public. However, the quantity of material is beyond the scope of an individual to meaningfully make sense of. To effectively organise this material a bespoke toolset is needed to analyse this collection of audio and determine 1. Where there are relationships between segments of audio files, and 2. What the differences are between two related audio recordings The field of audio data analysis is a large and highly commercialised one. Toolsets exist that can comprehend connections between large audio datasets range for compositional use (eg: FluCoMa, CataRT) and information retrieval (MIR, propriatry algorithms for applications such as Shazam and Spotify). However, existing commercial and non*commercial tools are not developed for the analysis of tape archives. Existing toolsets are not able to effectively analyse collections of this size and type. Such tools are developed to compare perfect digital audio files and the process of digitisation from historical analogue materials to digital audio files unavoidably introduces significant variance to the signal (unpredictable fluctuations in pitch, tone saturation, frequency filtering). This means that two instances of the same recording can fail to be recognised as related when using existing tools. TAAT will address this shortcoming and find new strategies for recognising relationships between digitised audio files. Over 12 months of dedicated development, a number of approaches will be implemented and tested - including signal comparison within the time and frequency domain, Music Information Retrieval analysis, and neural networks - using materials from the existing large audio databases in the Roberto Gerhard and Ernest Berk tape archives. Methods will be tested and modified in order to ascertain the most effective path forward to develop new processes for the analysis of large audio archives derived from analogue sound sources. The project team will include a musicologist who will apply the TAAT to new contexts and provide feedback to the project team. This project will act as a trial for the development of future research in both digital archive management and the creation of new musicological knowledge in the field of England's electronic music history and to make sense of the masses of recorded materials in archive collections around the world.