Speech recognition
Speech recognition is the computational conversion of spoken language from audio into a machine-readable representation, typically text.
Effective Conversational AI
Speech-to-text, or automatic speech recognition, converts spoken audio into written text.
- How it is used
- Converts callers’ speech into text for a chatbot classifier to interpret during voice interactions.
- How deep it goes
- Explains transcription failures, accuracy checks, and practical workarounds, but does not describe speech-recognition model internals.
- Examples
- Uses “residential” becoming “presidential,” “VIN” becoming “BIN,” and callers speaking or keying in member IDs.
- What you can do after
- Diagnose/debug with it by checking transcription accuracy, preserving recurring errors in tests, and accommodating valid caller responses.
Read how Effective Conversational AI explains speech recognition →
The Complete Obsolete Guide to Generative AI
Speech recognition is the conversion of spoken audio into a machine-readable representation of language.
- How it is used
- Uses speech recognition to turn recordings into searchable transcripts for summaries and targeted questions.
- How deep it goes
- Demonstrates a service workflow and a runnable Whisper transcription call, without detailing recognition-model internals.
- Examples
- Uses missed video conferences, a YouTube-video summary, generated narration, and a MyAudio.flac transcription script.
- What you can do after
- Apply it via OpenAI Whisper or transcription services to create text transcripts and feed them into summary workflows.
Read how The Complete Obsolete Guide to Generative AI explains speech recognition →
Microsoft Azure in Action
Speech to text is the automatic conversion of spoken language in an audio signal into written text.
- How it is used
- Turns recorded interviews and calls into searchable transcripts for archives, accessibility, quality assurance, and reuse.
- How deep it goes
- Explains resource provisioning and an asynchronous batch workflow through submission, timestamps, status checks, and output retrieval; does not detail recognition internals.
- Examples
- Uses recorded author interviews, an Azure CLI resource-creation command, and REST requests for batch submission and status checks.
- What you can do after
- Operate/configure it by provisioning Azure Speech and submitting and checking a batch transcription job for recorded audio.
Read how Microsoft Azure in Action explains speech recognition →
Data Analysis with LLMs
Speech recognition is the automatic conversion of spoken language into a machine-readable representation, commonly text.
- How it is used
- Turns recorded speech into searchable text and supports voice-driven applications.
- How deep it goes
- Explains Whisper’s transcription endpoint, file handling, and returned text, without detailing recognition internals.
- Examples
- Uses recorded employee meetings and a short sample audio clip to demonstrate searchable and transcribed speech.
- What you can do after
- Apply transcription through OpenAI’s Python library to turn an audio-file path into printed text.
Read how Data Analysis with LLMs explains speech recognition →
Build Applications with Local AI Models on a Mac
The conversion of spoken audio into written text.
- How it is used
- Converts microphone audio into text that a local chat application can pass to an LLM.
- How deep it goes
- Defines ASR, outlines Whisper’s encoder-decoder process, and gives audio requirements and a transcription call; it does not derive model internals.
- Examples
- Uses a microphone transcription script and a browser chat interface that turns recorded speech into an LLM prompt.
- What you can do after
- Build it from scratch with Python and MLX Whisper to transcribe microphone recordings and feed their text into a local chat application.
Read how Build Applications with Local AI Models on a Mac explains speech recognition →
Data Without Labels
Speech recognition is the computational conversion or identification of spoken language from an audio signal.
- How it is used
- Turns voice commands and recorded customer calls into text data for analysis and service improvement.
- How deep it goes
- Connects automatic speech recognition to deep learning and outlines DBN pretraining and fine-tuning, but omits implementation details.
- Examples
- Uses Alexa, Siri, Translator, Baidu, and transcribed call-center recordings as concrete speech-related applications.
- What you can do after
- Evaluate speech recognition as a deep-learning application and identify recorded calls as a source of text data.
Related concepts
- pipelinepart of · A pipeline is an ordered or coordinated sequence of connected processing stages or tasks that passes data, work, or results toward an outcome.
- MLX Whisperuses · MLX Whisper is a speech-recognition implementation that packages the Whisper model in a format optimized for Apple's MLX framework.
- Whisperimplemented by · Whisper is a family of OpenAI automatic speech-recognition models that transcribe and translate spoken audio, including multilingual audio.
- word error ratemeasured by · Word error rate is the ratio of word-level substitutions, deletions, and insertions required to transform a recognized transcription into a reference transcription.
- action itemproduces · An action item is a specific task or follow-up commitment identified or assigned for completion.
- speaker notespart of · Supplementary text associated with presentation slides that guides a speaker during delivery.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.