Product Glossary
This glossary defines the terms used across Capacity Private Cloud documentation. Terms are grouped by product area — Speech and Voice Biometrics — and listed alphabetically within each section.
Speech
Acoustic Model — A database of sounds of a language and what text they translate to — the sound-to-text relationship for the language. An ASR engine compares incoming audio to this model and outputs words built from the sounds. These are strings of letters treated as a first draft for the words; the final output is formed when the SLM prediction or grammar rules are applied.
Automatic Speech Recognition (ASR) — Software that converts human speech audio into text. ASR runs in real-time streaming mode, typically using predefined grammars, and in offline/batch mode, typically for long, free-form transcription based on statistical language models (SLM). The platform's ASR uses a Deep Neural Network (DNN) engine.
Call Progress Analysis (CPA) — Software that uses Voice Activity Detection (VAD) and Answering Machine Detection (AMD) to distinguish machines from humans, and business lines from residential lines. CPA accurately connects human responders to agents or delivers messages to voicemail, then informs the outbound messaging application what to do next — hand off to a live agent or leave a personalized voicemail.
Call Summarization — The process of automatically generating a concise summary of the content and key points discussed during a phone call or voice conversation, based on the transcribed text.
Continuous Transcription — In continuous mode the engine processes incoming audio a chunk at a time. The VAD detects when speech ends, and the audio processing performs word prediction — first pass, final pass, and post-processing — for every chunk continuously without waiting for the end of the audio.
Directed Dialog — A specification of what the caller will be asked and what the system's options are based on all possible responses. Directed dialogs follow specified paths until the end of the dialog.
Enhanced Transcription — A transcription mode that combines transcription with grammars. This enables NLP-style tasks such as slot filling and entity recognition.
Grammar — A set of rules that models what users are expected to say. A grammar acts as a filter, limiting the comparison and constraining the acoustic search when ASR tries to identify what a user said. In addition to providing the set of words users are expected to speak, a grammar also provides semantic interpretation — for example, it can interpret all items in a group like {"yes", "yeah", "ok", "of course", "sure", "that's fine", "got it"} as =YES.
Grammar-Based ASR — Automatic speech recognition based on the closed-set rules of a grammar.
gRPC — A cross-platform, open-source, high-performance Remote Procedure Call (RPC) framework, originally created by Google. This is the preferred and recommended protocol for communication with the platform. gRPC runs in any environment and efficiently connects services within and across data centers with support for load balancing, tracing, health checking, and authentication. It is optimized for scalability and is cloud- and microservices-native. All major programming languages can access the platform's API over gRPC.
Interactive Voice Response (IVR) — A system with automated menu choices that uses both ASR and TTS to interact with human callers. The system prompts for a choice, the user says their choice, and the system responds and acts accordingly.
Language Identification — Also known as language detection or language recognition, the process of determining the language of a given piece of text or spoken content.
Language Translation — The process of converting text or spoken content from one language (the source language) into another language (the target language) while preserving the meaning and context of the original message.
MRCP — Media Resource Control Protocol, a communication protocol between applications (voice IVR platforms) and the ASR and TTS resources serving them. Primarily used by voice application platforms, call centers, call recording providers, telephony trunks, and network switches, using W3C standards. While the preferred approach is the gRPC-based Capacity Private Cloud API for speech and voice services, MRCP remains prevalent in customer service platforms. The platform supports all major voice platforms and IVRs using MRCP. There are two main versions in use: MRCPv1 and MRCPv2.
Natural Language Understanding (NLU) — A subfield of artificial intelligence and computational linguistics focused on a computer system's ability to understand and interpret human language as it is naturally spoken or written. Unlike simple keyword matching, NLU involves comprehending the nuances, context, and intent behind user inputs. The platform's NLU stack includes intent and slot extraction (see Neuron), Call Summarization, Sentiment Analysis, Language Identification, and Language Translation.
Neural TTS — Text-to-speech generated using deep neural network (DNN) models, producing more natural-sounding output than rule-based synthesis. See also Text to Speech (TTS).
Neuron — Capacity's intent and slot extraction service within the NLU stack. Identifies what a speaker means (the intent) and the parameters that qualify the intent (the slots) from transcribed input, so downstream applications can act on intent rather than raw transcription.
Sentiment Analysis — Also known as opinion mining, the process of identifying, extracting, and quantifying subjective information from text. Sentiment analysis determines the emotional tone, attitude, or sentiment expressed in a piece of text, typically categorized as positive, negative, or neutral.
Speaker Diarization — The process of partitioning an audio recording into segments based on who is speaking when. Diarization identifies and labels different speakers within an audio stream — essential for understanding multi-speaker environments such as meetings, interviews, and call centers.
Speech Synthesis Markup Language (SSML) — A standardized markup language used to control various aspects of speech synthesis (the artificial production of human speech). SSML is part of the W3C standards and is commonly used in text-to-speech (TTS) systems to control prosody, pronunciation, and word emphasis.
Statistical Language Model (SLM) — A model of an entire language that specifies relationships between words.
Text to Speech (TTS) — Software that converts text into audio. The audio is then played back by the client application, responding to end users with spoken words. Each supported language comes with a choice of male or female voices. See also Neural TTS for deep-neural-network-based synthesis.
Training an Acoustic Model — Running thousands of hours of audio with transcribed speech through automated training, which compares transcripts to sounds. The training process captures how words are pronounced by a diverse set of people and builds the database of sound-to-text mappings.
Training a Language Model — Running millions of lines of text through automated training, which captures word relationships and builds a database of word-relationship probabilities.
Transcription ASR — Automatic speech recognition based on the open rules of a statistical language model (SLM) rather than a closed-set grammar.
Voice Biometrics
Authentication / Verification — The process of verifying the identity of a person using their voiceprint. This typically involves matching the individual's live voice input against a stored voiceprint.
Biometric Model — A biometric Deep Neural Network (DNN) model, an advanced machine learning architecture used to perform identification, authentication, and verification. A custom model is often required for specific text-dependent phrases.
Channel Playback Detection — Detects that audio is being played back through a speaker device rather than spoken live. Used in active, passive, and hybrid use cases.
Enrollment — The process of recording a person's voice to create a voiceprint. During enrollment, the user's voice features are captured and stored for future comparison.
False Accept Rate (FAR) — The rate at which an unauthorized person is incorrectly accepted by the voice biometric system as an authorized user.
False Reject Rate (FRR) — The rate at which an authorized user is incorrectly rejected by the voice biometric system.
Footprint Playback Detection — Identifies that an audio file has already been received and processed by the voice biometric engine. Operates on the premise that a human never repeats a passphrase in exactly the same way. Used in active verification use cases.
Liveness Detection — A method that confirms voice input comes from a live human rather than a recording or synthetic imitation. Typically done by prompting the speaker to say a randomly chosen phrase — for example, "An apple a day keeps the doctor away."
Playback Detection — Determines whether recorded or live audio is being played back into the production system.
Speaker Identification — The process of determining which registered speaker's voiceprint matches a given voice input from a group of possible speakers.
Synthetic Speech Detection — Identifies that the audio being recorded was generated by a speech synthesizer (text-to-speech or voice-conversion deepfakes). Supported across both passive and active authentication flows.
Text-Dependent Verification — Voice verification in which the speaker must say a specific phrase — for example, "My voice is my password." The system compares the spoken input against the pre-recorded version of the same phrase.
Text-Independent Verification — Voice verification that does not require the speaker to say a specific phrase. The system verifies the speaker based on any spoken input.
Voice Biometrics — Technology for recognizing a person based on the unique characteristics of their voice, typically used for authentication and verification.
Voiceprint — A digital model representing the unique characteristics of an individual's voice, used for comparison during authentication.