FEARLESS STEPS Challenge (FS-2): Supervised Learning with Massive\n Naturalistic Apollo Data

The Fearless Steps Initiative by UTDallas-CRSS led to the digitization,\nrecovery, and diarization of 19,000 hours of original analog audio data, as\nwell as the development of algorithms to extract meaningful information from\nthis multi-channel naturalistic data resource. The 2020 FEARLESS STEPS (FS-2)\nChallenge is the second annual challenge held for the Speech and Language\nTechnology community to motivate supervised learning algorithm development for\nmulti-party and multi-stream naturalistic audio. In this paper, we present an\noverview of the challenge sub-tasks, data, performance metrics, and lessons\nlearned from Phase-2 of the Fearless Steps Challenge (FS-2). We present\nadvancements made in FS-2 through extensive community outreach and feedback. We\ndescribe innovations in the challenge corpus development, and present revised\nbaseline results. We finally discuss the challenge outcome and general trends\nin system development across both phases (Phase FS-1 Unsupervised, and Phase\nFS-2 Supervised) of the challenge, and its continuation into multi-channel\nchallenge tasks for the upcoming Fearless Steps Challenge Phase-3.\n