From 6346954e7ad612800e766821db0c6cecc2f64b82 Mon Sep 17 00:00:00 2001
From: Dave Horton <daveh@beachdognet.com>
Date: Thu, 18 Aug 2022 23:18:24 +0200
Subject: [PATCH] session-level speech hints, strip trailing punctuation on
 continuous asr (#151)

---
 lib/session/call-session.js | 13 +++++++++++++
 lib/tasks/config.js         |  7 +++++++
 lib/tasks/gather.js         | 20 ++++++++++++++++++--
 lib/tasks/transcribe.js     | 12 ++++++++++--
 4 files changed, 48 insertions(+), 4 deletions(-)

diff --git a/lib/session/call-session.js b/lib/session/call-session.js
index d2880fe8..12341972 100644
--- a/lib/session/call-session.js
+++ b/lib/session/call-session.js
@@ -261,6 +261,19 @@ class CallSession extends Emitter {
 
   get recordState() { return this._recordState; }
 
+  set globalSttHints({hints, hintsBoost}) {
+    this._globalSttHints = {hints, hintsBoost};
+  }
+
+  get hasGlobalSttHints() {
+    const {hints = []} = this._globalSttHints || {};
+    return hints.length > 0;
+  }
+
+  get globalSttHints() {
+    return this._globalSttHints;
+  }
+
   async notifyRecordOptions(opts) {
     const {action} = opts;
     this.logger.debug({opts}, 'CallSession:notifyRecordOptions');
diff --git a/lib/tasks/config.js b/lib/tasks/config.js
index ea4964dc..230db0d3 100644
--- a/lib/tasks/config.js
+++ b/lib/tasks/config.js
@@ -87,6 +87,13 @@ class TaskConfig extends Task {
         cs.asrTimeout = this.recognizer.asrTimeout;
         cs.asrDtmfTerminationDigit = this.recognizer.asrDtmfTerminationDigit;
       }
+      if (Array.isArray(this.recognizer.hints)) {
+        const obj = {hints: this.recognizer.hints};
+        if (typeof this.recognizer.hintsBoost === 'number') {
+          obj.hintsBoost = this.recognizer.hintsBoost;
+        }
+        cs.globalSttHints = obj;
+      }
       this.logger.info({
         recognizer: this.recognizer,
         isContinuousAsr: cs.isContinuousAsr
diff --git a/lib/tasks/gather.js b/lib/tasks/gather.js
index 95ccca13..60783972 100644
--- a/lib/tasks/gather.js
+++ b/lib/tasks/gather.js
@@ -121,6 +121,13 @@ class TaskGather extends Task {
     await super.exec(cs);
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
+    if (cs.hasGlobalSttHints) {
+      const {hints, hintsBoost} = cs.globalSttHints;
+      this.hints = this.hints.concat(hints);
+      if (!this.hintsBoost && hintsBoost) this.hintsBoost = hintsBoost;
+      this.logger.debug({hints: this.hints, hintsBoost: this.hintsBoost},
+        'Gather:exec - applying global sttHints');
+    }
     if (!this.isContinuousAsr && cs.isContinuousAsr) {
       this.isContinuousAsr = true;
       this.asrTimeout = cs.asrTimeout * 1000;
@@ -286,7 +293,7 @@ class TaskGather extends Task {
       ].forEach((arr) => {
         if (this[arr[0]]) opts[arr[1]] = true;
       });
-      if (this.hints.length > 1) {
+      if (this.hints.length > 0) {
         opts.GOOGLE_SPEECH_HINTS = this.hints.join(',');
         if (typeof this.hintsBoost === 'number') {
           opts.GOOGLE_SPEECH_HINTS_BOOST = this.hintsBoost;
@@ -333,7 +340,7 @@ class TaskGather extends Task {
           'AZURE_REGION': this.sttCredentials.region
         });
       }
-      if (this.hints && this.hints.length > 1) {
+      if (this.hints && this.hints.length > 0) {
         opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
       }
       if (this.altLanguages && this.altLanguages.length > 0) {
@@ -487,6 +494,15 @@ class TaskGather extends Task {
       }
       if (this.isContinuousAsr) {
         /* append the transcript and start listening again for asrTimeout */
+        const t = evt.alternatives[0].transcript;
+        if (t) {
+          /* remove trailing punctuation */
+          if (/[,;:\.!\?]$/.test(t)) {
+            this.logger.debug('TaskGather:_onTranscription - removing trailing punctuation');
+            evt.alternatives[0].transcript = t.slice(0, -1);
+          }
+          else this.logger.debug({t}, 'TaskGather:_onTranscription - no trailing punctuation');
+        }
         this.logger.info({evt}, 'TaskGather:_onTranscription - got transcript during continous asr');
         this._bufferedTranscripts.push(evt);
         this._clearTimer();
diff --git a/lib/tasks/transcribe.js b/lib/tasks/transcribe.js
index c47c02fc..c06e209b 100644
--- a/lib/tasks/transcribe.js
+++ b/lib/tasks/transcribe.js
@@ -62,6 +62,14 @@ class TaskTranscribe extends Task {
     super.exec(cs);
     const {updateSpeechCredentialLastUsed} = require('../utils/db-utils')(this.logger, cs.srf);
 
+    if (cs.hasGlobalSttHints) {
+      const {hints, hintsBoost} = cs.globalSttHints;
+      this.hints = this.hints.concat(hints);
+      if (!this.hintsBoost && hintsBoost) this.hintsBoost = hintsBoost;
+      this.logger.debug({hints: this.hints, hintsBoost: this.hintsBoost},
+        'Transcribe:exec - applying global `sttHints');
+    }
+
     this.ep = ep;
     this.ep2 = ep2;
     if ('default' === this.vendor || !this.vendor) this.vendor = cs.speechRecognizerVendor;
@@ -158,7 +166,7 @@ class TaskTranscribe extends Task {
       ].forEach((arr) => {
         if (this[arr[0]]) opts[arr[1]] = true;
       });
-      if (this.hints.length > 1) {
+      if (this.hints.length > 0) {
         opts.GOOGLE_SPEECH_HINTS = this.hints.join(',');
         if (typeof this.hintsBoost === 'number') {
           opts.GOOGLE_SPEECH_HINTS_BOOST = this.hintsBoost;
@@ -218,7 +226,7 @@ class TaskTranscribe extends Task {
         'AZURE_SUBSCRIPTION_KEY': this.sttCredentials.api_key,
         'AZURE_REGION': this.sttCredentials.region
       });
-      if (this.hints && this.hints.length > 1) {
+      if (this.hints && this.hints.length > 0) {
         opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
       }
       if (this.altLanguages.length > 1) opts.AZURE_SPEECH_ALTERNATIVE_LANGUAGE_CODES = this.altLanguages.join(',');