# Automating language detection and sentiment analysis in Kustomer

> Kustomer can automatically detect the language of inbound messages and analyze their sentiment These capabilities help teams understand customer intent route 

Source: https://support.kustomer.com/en_us/automating-language-detection-and-sentiment-analysis-in-kustomer-H1a4G1KFZe

Last updated: 2026-03-06T23:35:21.095Z

Kustomer can automatically detect the language of inbound messages and analyze their sentiment. These capabilities help teams understand customer intent, route conversations appropriately, and support message translation.

Language detection and sentiment analysis run automatically when the **Intelligence app** is installed.

### In this article

*   [Understanding language detection and sentiment analysis](#understanding)
    
*   [Installing the Intelligence app](#installing)
    
*   [Using language detection](#using)
    
*   [Viewing sentiment data](http://viewing)
    

### Understanding language detection and sentiment analysis

Language detection and sentiment analysis run through the **detect-language-at-message-creation** workflow installed by the _Intelligence app_.

When a new inbound message is created, the workflow performs two steps:

1.  **Detect the most likely language** of the message and calculate a confidence score.
    
2.  **Analyze the message sentiment** and assign a sentiment classification with a score.
    

The workflow runs automatically for all inbound messages that meet these conditions:

*   The message is not imported.
    
*   The message contains text.
    

If the detected language meets the confidence threshold (**minimum language score ≥ 0.6**), Kustomer updates the relevant metadata.

The following attributes can also be updated:

*   Message language
    
*   Customer language
    
*   Conversation language
    

Language detection and sentiment analysis are powered by Kustomer’s modern NLP pipeline.

The system uses:

*   open-source machine learning models
    
*   Large language models (LLMs)
    
*   Amazon Comprehend as a fallback when errors occur
    

This layered approach improves accuracy and reliability across a wide range of languages and message types.

### Installing the Intelligence app

The Intelligence app is required for language detection and sentiment analysis. 

To install the Intelligence app:

1.  From the left navigation panel, click **Apps**. The **Apps** page opens. 
    
2.  Search for **Intelligence** in the Kustomer app directory. Click the app to open its application page. 
    
3.  At the top left of the application's page, click **Install** to install the application. 
    

With the app installed, you must enable language detection and sentiment analysis. 

1.  From the left navigation within any **Apps** page, click **Intelligence**. 
    
2.  Check the box to **Detect language automatically**. 
    

### Using language detection

Detected language helps power several product capabilities.

#### Message translation

Language detection enables **two-way message translation** between customers and agents. When translation is enabled, Kustomer automatically identifies the original message language and translates messages as needed.

#### Conversation metadata

Detected language is stored as metadata and displayed in:

*   Conversation attributes
    
*   Message attributes
    

This metadata can be used for:

*   Reporting
    
*   Routing workflows
    
*   Automation rules
    
*   Analytics
    

### Viewing sentiment data

Kustomer analyzes message sentiment and surfaces the result in a sentiment insight card, available in the **Insights panel** from any opened conversation. 

Sentiment analysis assigns a score and classification that indicate the overall tone of the message.

Examples include:

*   Positive
    
*   Neutral
    
*   Negative
    

These insights help teams identify frustrated customers or high-risk conversations more quickly.

Sentiment scores represent the model’s confidence in the detected tone of the message. They are displayed as percentages for easier interpretation in the interface.[Understanding language detection and sentiment analysis](#understanding)
