From b4fd26a42a27b0ba7d10213317ae11a12493c089 Mon Sep 17 00:00:00 2001 From: el-tocino Date: Thu, 1 Aug 2019 10:33:03 -0500 Subject: [PATCH] minor updates --- not-wake-words/README.md | 7 +------ not-wake-words/noises/README.md | 8 +++++++- 2 files changed, 8 insertions(+), 7 deletions(-) diff --git a/not-wake-words/README.md b/not-wake-words/README.md index 9b1689c..8b7ecd1 100644 --- a/not-wake-words/README.md +++ b/not-wake-words/README.md @@ -1,9 +1,4 @@ ## Not wake words -When uploading files to this directory, include a metadata.csv file. The format of the metadata file should follow: - -``` -filename, transcript, language, related wake word (optional) -``` - +This directory contains subdirectories for language-specific clips and a general noises. See each language directory for information on adding clips to those directories. The noises subdirectory is meant only for non-verbal or unintelligible audio clips. diff --git a/not-wake-words/noises/README.md b/not-wake-words/noises/README.md index d7042cb..d5fd462 100644 --- a/not-wake-words/noises/README.md +++ b/not-wake-words/noises/README.md @@ -1,3 +1,9 @@ ## Noises -This directory is meant to house any clips that contain sound, but not recognizable speech of any language. Loud noises, conversations that can't be clearly discerned, sirens, bangs, cracks, coughs, and so on. +This directory is meant to house any clips that contain sound, but not recognizable speech of any language. Loud noises, conversations that can't be clearly discerned, sirens, bangs, cracks, coughs, and so on. Use a metadata.csv file with the following format when uploading clips: + +``` +filename, short description of noise +``` + +While not every clip will have a description, it is helpful to know if targeting a certain type of noise when building models.