Tone analytics (Classifications)

Tone analytics is currently available for English and French languages only, as indicated in the language support topic.

Tone analysis is done by using a pre-built classifications model, which provides an easy way to detect language tones in written text. It detects seven tones: sad, frustrated, satisfied, excited, polite, impolite, and sympathetic.

Analyzing tone

To detect tone, use the language-specific classifications model ID in your API request.

The language-specific tone model ID is formatted as tone-classifications-xx-v1, where xx is a two-character language code. Languages available include:

Language	Code
English	`en`
French	`fr`

Example parameters.json file:

{
  "language": "en",
  "text": "This is example text in English.",
  "features": {
    "classifications": {
      "model": "tone-classifications-en-v1"
    }
  }
}

Example cURL request:

curl --request POST \
--header "Content-Type: application/json" \
--user "apikey":"{apikey}" \
"{url}/v1/analyze?version=2021-08-01" \
--data @parameters.json

Understanding tone analytics

The model returns scores for the following tones:

Tone	Description
`excited`	Showing personal enthusiasm and interest
`frustrated`	Feeling annoyed and irritable
`impolite`	Being disrespectful and rude
`polite`	Displaying rational, goal-oriented behavior
`sad`	An unpleasant passive emotion
`satisfied`	An affective response to perceived service quality
`sympathetic`	An affective mode of understanding that involves emotional resonance

Example response:

{
"usage": {
  "text_units": 1,
  "text_characters": 60,
  "features": 1
},
"language": "en",
"classifications": [
  {
    "confidence": 0.564849,
    "class_name": "excited"
  },
  {
    "confidence": 0.355816,
    "class_name": "satisfied"
  },
  {
    "confidence": 0.126127,
    "class_name": "polite"
  },
  {
    "confidence": 0.026995,
    "class_name": "sympathetic"
  },
    {
    "confidence": 0.012211,
    "class_name": "frustrated"
  },
  {
    "confidence": 0.011065,
    "class_name": "sad"
  },
  {
    "confidence": 0.000872,
    "class_name": "impolite"
  }
]
}

Migrating from Watson Tone Analyzer Customer Engagement endpoint to Natural Language Understanding

You can migrate your Watson Tone Analyzer customer-engagement analysis requests to Natural Language Understanding. This can help you better understand your interactions with customers and improve your communications generally, or for specific customers.

Reformatting your input data

In Watson Tone Analyzer, you pass the /v3/tone_chat method a JSON ToneChatInput object consisting of utterances, text, and an optional user string fields. For Natural Language Understanding, you pass a JSON object that contains text to be analyzed, and a language-specific model classification ID, to the /v1/analyze method.

Tone Analyzer request mapping to NLU .

Understanding your response content

In Watson Tone Analyzer, the service returns a JSON UtteranceAnalyses object that contains a single field, utterances_tone, which contains an array of UtteranceAnalysis objects, including score and tone_id. For Natural Language Understanding, a classifications field is returned, containing confidence and class_name objects that correspond to the Watson Tone Analyzer score and tone_id objects.

Tone Analyzer response mapping to NLU .