Understanding › Sentiment analysis

Sentiment analysis

Analyze tone, emotions, and sentiment trends from transcribed audio — with Arabic sentiment interpretation that's aware of cultural context. Transcribe first, then point this endpoint at the transcriptionId.

What you get

Sentiment analysis processes already transcribed audio to extract emotional insights and sentiment patterns, at multiple depth levels.

Insight	Detail
Overall sentiment	Sentiment and polarity analysis across the full content.
Key emotions	Detected emotions with confidence scoring.
Speaker breakdowns	Speaker-level sentiment for each participant.
Temporal trends	How sentiment shifts across the conversation.

Typical use cases: call center quality and customer experience monitoring; public speech, media, and interview sentiment tracking; meeting intelligence and post-call emotional analysis.

This is post-call analysis of an existing transcription. For real-time per-turn sentiment during a live call, use the Sentiment event on WS /api/v1/listen — no extra request needed. Note that confidence_score here is the confidence of the sentiment analysis itself, unrelated to ASR confidence.

How it works

This endpoint runs on top of an existing transcription — it never touches raw audio itself.

Step	What happens
1 · Transcribe first	Run the audio through Audio Transcription.
2 · Use the transcription ID	Send the returned `transcriptionId` to Sentiment Analysis.
3 · Choose depth	Pick `light`, `standard`, or `deep` depending on how detailed you want the output.
4 · Review insights	Receive emotional and sentiment analysis for the full content and individual speakers.

Want sentiment per speaker turn, in one call? Diarization + sentiment transcribes, splits speakers and scores sentiment in a single request.

Endpoint

POST /api/v1/audio/{transcriptionId}/sentiment-analysis

Authenticated with the x-api-key header.

Parameter	In	Type	Required	Description
`transcriptionId`	path	string	Yes	ID from the Audio Transcription response.
`analysis_depth`	body	string	No	`light`, `standard`, or `deep`.

Analyze a transcription

Replace 805059bf-7c3f-4a1e-9d2b-1f0c6ae83b47 with the transcriptionId your transcription call returned.

curl

curl -X POST "https://api.munsit.com/api/v1/audio/805059bf-7c3f-4a1e-9d2b-1f0c6ae83b47/sentiment-analysis" \ -H "x-api-key: YOUR_MUNSIT_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "analysis_depth": "standard" }'

Response

The response covers full-content sentiment plus per-speaker and per-moment detail. Depth controls how much of it is populated.

Field	What it holds
`language`	Detected language of the analyzed transcript.
`overall_sentiment`	Overall sentiment and polarity for the full content.
`key_emotions`	Key emotions detected, with confidence scoring.
`speaker_sentiment`	Speaker-level sentiment breakdowns.
`sentiment_trends`	Temporal sentiment trends across the conversation.
`critical_moments`	Notable emotional moments in the conversation.
`confidence_score`	Confidence for the analysis as a whole.
`analysis_depth`	The depth level the analysis ran at.
`processing_metadata`	Metadata about the processing run.

Go further

Sentiment is one lens on a transcript — the others are a call away.

Transcribe audioProduce the transcriptionId this endpoint needs.POST /speech-to-text Diarization + sentimentSpeakers and sentiment together, in one request.POST /stt-diarization Keyword extractionPull entities, terms and themes from the same transcript.POST …/keyword-extraction TranslationTranslate the transcript with streaming output.POST /translation/stream

Working with an AI assistant? Every page is available as Markdown: add .md to the URL, or send an Accept: text/markdown header. For the whole documentation in one request, point it at llms-full.txt; the page index is llms.txt. Or use Copy Page, top right.