75 Speech Recognition Features

If you are considering the addition of Speech-to-text (or automatic speech recognition (ASR)) to your product, there are several providers with similar features & service options. While features are important it's critical to choose what fits best to your use case. Read my blog covering?10 factors to consider when evaluating a speech recognition engine.

In this blog, I will cover 75 key product features of speech recognition systems that are available in production or limited release already. I have categorized the features and options below:

1. Input:

• speech to text
• PSTN/SIP support
• speech input
• multi-channel support
• dialect support
• real-time transcription
• offline mode
• file formats

2. Output:

• text output
• speech output
• custom responses
• response time
• interim and final transcription
• transcription export in VTT, SRT, etc
• paragraph formatting
• natural language generation/text to speech
• throughput/processing time
• latency
• pre-built user interface and integrations

3. Core Speech Recognition:

• accuracy
• Word Error Rate (WER)
• context awareness
• far-field speech recognition
• keyword spotting
• Custom Vocabulary
• PII detection, removal, masking
• automatic punctuation & casing
• word-level confidence score
• word-level timestamps
• profanity detection
• dictation mode
• conversation mode
• data collection option (on/off)
• music and special audio detection
• short form speech (IVR, etc)

4. Audio Intelligence:

• diarization
• speaker identification
• noise detection and reduction
• silence detection
• voice biometrics
• language detection
• translation
• natural language query

5. Conversation Quality Metrics:

• speech recognition accuracy
• translation accuracy
• speech quality
• filler words handling (detection, removal)
• talk time / listen time
• words per minute by speaker
• interruptions and overlaps
• latency
• jitter

6. Conversation Intelligence & Insights:

• intent recognition
• sentiment analysis
• topic detection
• extract summary
• abstract or executive summary
• content safety detection
• tone detection
• action or to do classification
• question detection
• answer detection
• content redaction
• conversation classification or categorization (example reason for call)
• event identification
• key phrase detection

7. Customizations:

• domain adaptation
• acoustic model training
• language model training
• custom voice
• custom wake word
• custom vocabulary
• custom entity recognition
• custom classifiers
• personalized learning from transcript edits

If you want to learn more about these and get a copy of the features-and-vendor mapping, I have curated a side-by-side comparison for top speech recognition providers which I am happy to share. Let me know via your comments or?DMs. For further reading, please read my blog about the overview of speech recognition on?Medium.