Define Category

In the Category field, enter the category name.
Select the Audio File Type as Stereo or ‘Mono. For mono audio file type, speaker diarization option is available for selection. By default, option On is selected.
Note
System does not allow Admin to change the Speaker Diarization mode after a Business Rule is created.
Select option as ‘On’ or ‘Off’ in Sentiment Analysis field to analyze sentiment expressed at every turn in the conversation and also overall Agent conversation sentiment and Customer conversation sentiment. Sentiment can be enabled only for US-ENG and Hinglish calls since our models support only these languages out of the box.
Select any one of the Speaker Diarization engines from the drop-down list. This is to identify speakers as Agent or Customer at every turn of the conversation for Stereo Calls and for Mono Calls which are diarized.
- Model 3 (Nuance)
- Model 4 (Deep Learning)
Select the BR Evaluation Engine as Transcriber.

Select any of the below Transcription Engine from the drop-down list:

If you select FANO ASR engine, you can choose only Cantonese-China(zh) language as Agent and Customer language. This Engine supports only Cantonese language.
FANO ASR is not available for all the customers. This option is provided based on delivery-driven engagement.
If you select UNIPHORE ASR engine, you can choose only Vietnamese or Bahasa-Indonesia language as agent and customer language. This engine supports only Vietnamese and Bahasa-Indonesia languages.

If you select UNIPHORE ASR - GPU, you can choose any of the language that follows as agent or customer language:

1. English-USA	2. English-Indian	3. English-Great Britain	4. Hinglish
5. English-Filipino	6. Arabic	7. German	8. Japanese
9. French- Europe	10. Tamil	11. Malayalam	12. Tagalog
13. Spanish -North America	14. Kannada	15. English-Australasia	16. French Canadian
17. Vietnamese	18. Gujarati (Introduced in version 2023.4)	19. Spanish-Europe (Introduced in version 2023.4)	20. Marathi (Introduced in version 2023.4)
21. Bengali (Introduced in version 2024.1)

These languages are available in the Agent Language and Customer Language drop downs in an alphabetical (A-Z) order.

If you select Transcription Engine as “NTE”, you have an option to enable the Language Detection feature.

Select Language Identification Model (NTE LID or Uniphore LID) from the drop-down list. Uniphore’s Language Identification Model automatically identifies the language of the Speaker. Currently, U-Analyze supports English, Cantonese, Mandarin language detection and flags anything other than these languages as Others. The identified language can be used as the Language Pack for Transcription of the Audio.
Select Language Identification Model as NTE from the drop-down list. The different configurations for NTE Language Identification Model are as follows:
1. NTE Mono with Speaker Diarization ON and Language Detection ON.
2. NTE Mono with Speaker Diarization ON and Language Detection OFF.
3. NTE Mono with Speaker Diarization OFF and Language Detection ON.
4. NTE Mono with Speaker Diarization OFF and Language Detection ON.
5. NTE Stereo with Language Detection ON.
6. NTE Stereo with Language Detection OFF.

Select the Language Detection checkbox to detect languages for calls which are processed through Transcription Engine (Nuance Transcription Engine) and for specific offsets in the calls.
Select Language Identification Model as NTE.
Select Agent Language from the drop-down list. Multiple Languages can be selected for each Agent.
Note
In Customer Language section, the languages which are selected in Agent Language drop-down list are automatically displayed.
Enter the first “n” seconds or last “n” seconds for language detection. A maximum of 45 seconds are allowed.

Select Agent Language from the drop-down list. Only one language can be selected.
Select Customer Language from the drop-down list. Only one language can be selected.

Select the Language Detection checkbox.
Select Language from the drop-down list. Multiple languages can be selected for common channel.
Enter the first “n” seconds or last “n” seconds for language detection. Maximum of 45 seconds are allowed.

Select Language from the drop-down list. Only one language can be selected for common channel.

Select the Language Detection checkbox.
Select Language Identification Model as NTE.
Select Agent Language from the drop-down list. Multiple languages can be selected for each Agent.
Enter the first “n” seconds or last “n” seconds for language detection. A maximum of 45 seconds are allowed.
Select Customer Language from the drop-down list. Multiple languages can be selected for each customer.
Enter the first “n” seconds or last “n” seconds for language detection. A maximum of 45 seconds are allowed.

Select Agent Language from the drop-down list. Only one language can be selected.

Select Customer Language from the drop-down list. Only one language can be selected.

Refer the below table for NTE 4.6 supported languages:

Table 4. List of supported languages

Click Set Profiling Parameters to provide profiling URL (Domain specific language model URL) for Agent and Customer, if required.
In profiling, keyphrases which are domain specific and proper nouns are included in the Keyphrase list to improve transcription accuracy. For more information, refer “Installation Guide” to create domain LMs.
Note
Domain specific language model URL is case sensitive.
Click Add Category () button to save and add the category. A confirmation message is displayed as “Category added successfully".
Click Next button to save data and move to Define Users screen.

In this section:

Uniphore Customer Portal