ben.carrasco/jambonz-feature-server
1
0
Fork 0
mirror of https://github.com/jambonz/jambonz-feature-server.git synced 2025-12-20 08:40:38 +00:00
Code Issues Packages Projects Releases Wiki Activity

Feat/speechmatics (#932)

Browse Source 
* wip

* initial working version of speechmatics

* linting
This commit is contained in:
Dave Horton
2024-10-12 18:42:53 -04:00
committed by GitHub
parent 84b7456c2d
commit f84b3793e1
7 changed files with 175 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
lib/utils/transcription-utils.js
View File

@@ -1,7 +1,4 @@
const {
TaskName,
} = require('./constants.json');
const {TaskName} = require('./constants.json');
const stickyVars = {
google: [
'GOOGLE_SPEECH_HINTS',
@@ -100,6 +97,12 @@ const stickyVars = {
assemblyai: [
'ASSEMBLYAI_API_KEY',
'ASSEMBLYAI_WORD_BOOST'
],
speechmatics: [
'SPEECHMATICS_API_KEY',
'SPEECHMATICS_HOST',
'SPEECHMATICS_PATH',
'SPEECHMATICS_SPEECH_HINTS',
]
};
@@ -466,12 +469,37 @@ const normalizeAssemblyAi = (evt, channel, language) => {
}
],
vendor: {
name: 'ASSEMBLYAI',
name: 'assemblyai',
evt: copy
}
};
};
const normalizeSpeechmatics = (evt, channel, language) => {
const copy = JSON.parse(JSON.stringify(evt));
const is_final = evt.message === 'AddTranscript';
const words = evt.results?.filter((r) => r.type === 'word') || [];
const confidence = words.length > 0 ?
words.reduce((acc, word) => acc + word.alternatives[0].confidence, 0) / words.length :
0;
const alternative = {
confidence,
transcript: evt.metadata?.transcript
};
const obj = {
language_code: language,
channel_tag: channel,
is_final,
alternatives: [alternative],
vendor: {
name: 'speechmatics',
evt: copy
}
};
return obj;
};
module.exports = (logger) => {
const normalizeTranscription = (evt, vendor, channel, language, shortUtterance, punctuation) => {
@@ -499,6 +527,8 @@ module.exports = (logger) => {
return normalizeAssemblyAi(evt, channel, language, shortUtterance);
case 'verbio':
return normalizeVerbio(evt, channel, language);
case 'speechmatics':
return normalizeSpeechmatics(evt, channel, language);
default:
if (vendor.startsWith('custom:')) {
return normalizeCustom(evt, channel, language, vendor);
@@ -828,7 +858,8 @@ module.exports = (logger) => {
...(cobaltOptions.enableConfusionNetwork && {COBALT_ENABLE_CONFUSION_NETWORK: 1}),
...(cobaltOptions.compiledContextData && {COBALT_COMPILED_CONTEXT_DATA: cobaltOptions.compiledContextData}),
};
} else if ('assemblyai' === vendor) {
}
else if ('assemblyai' === vendor) {
opts = {
...opts,
...(sttCredentials.api_key) &&
@@ -836,7 +867,8 @@ module.exports = (logger) => {
...(rOpts.hints?.length > 0 &&
{ASSEMBLYAI_WORD_BOOST: JSON.stringify(rOpts.hints)})
};
} else if ('verbio' === vendor) {
}
else if ('verbio' === vendor) {
const {verbioOptions = {}} = rOpts;
opts = {
...opts,
@@ -855,7 +887,16 @@ module.exports = (logger) => {
...(verbioOptions.speech_incomplete_timeout &&
{VERBIO_SPEECH_INCOMPLETE_TIMEOUT: verbioOptions.speech_incomplete_timeout}),
};
} else if (vendor.startsWith('custom:')) {
}
else if ('speechmatics' === vendor) {
opts = {
...opts,
...(sttCredentials.api_key) && {SPEECHMATICS_API_KEY: sttCredentials.api_key},
...(sttCredentials.speechmatics_stt_uri) && {SPEECHMATICS_HOST: sttCredentials.speechmatics_stt_uri},
...(rOpts.hints?.length > 0 && {SPEECHMATICS_SPEECH_HINTS: rOpts.hints.join(',')}),
};
}
else if (vendor.startsWith('custom:')) {
let {options = {}} = rOpts.customOptions || {};
const {sampleRate} = rOpts.customOptions || {};
const {auth_token, custom_stt_url} = sttCredentials;
@@ -923,6 +964,6 @@ module.exports = (logger) => {
setChannelVarsForStt,
setSpeechCredentialsAtRuntime,
compileSonioxTranscripts,
consolidateTranscripts
consolidateTranscripts,
};
};