Long-form recognition builder

Prev Next

Available in Classic and VPC

Using the CLOVA Speech Builder, you can request and manage speech recognition tasks through a UI environment. Builders are provided per domain, and you can use the following menus within the Builder:

  • Task list menu: Register and manage recognition tasks, edit results, and export output
  • Keyword Boosting menu: Register keywords for words you want to boost in recognition probability, and manage registered keywords
  • Settings menu: View domain information
Note

Some features may not be displayed depending on the domain type.

Launch the Builder

To run the Builder:

Note

You must create a domain before using the Builder. For more information, see Create a domain.

  1. From the NAVER Cloud Platform console, navigate to i_menu > Services > AI Services > CLOVA Speech.
  2. Click the Domain menu for the long-form recognition type.
  3. Click [Launch Builder] button for the domain you want to use.

Manage recognition tasks

From the Task list menu in the Builder, you can submit recognition tasks and monitor their progress. You can also edit recognition results using the Result editor and export them in the format you want.

Task list interface

The Task list menu includes the following components:
clovaspeech-builder_list2_ko

Component Description
① Menu name Shows the current menu name.
② Task management features Request task, delete, export, re-recognize.
③ Search Search tasks by media file name or result file location.
④ Task list Shows recognition tasks in the current view.

View recognition tasks

To view details about recognition tasks registered in the Builder:

  1. In the domain's builder, click the Task list menu.
  2. When the task list appears, review the recognition task details.
    • Target file: Name of the media file used for recognition
    • Language: Language used for the recognition task
    • Status: Current status of the task, displayed as one of the following:
      Status Description
      Waiting
      In progress
      Canceled
      Canceling
      Failed
      Completed
    • Request method: How the input file was uploaded
      • If the file was selected from Object Storage, click i-clovaspeech-window to open the Object Storage console.
    • Result file location: Object Storage path where the result file is saved
      • Click i-clovaspeech-window to open the Object Storage console.
    • Start date (UTC+09:00): Date and time when the task started
    • End date (UTC+09:00): Date and time when the task ended
    • Edit recognition result: Click [Edit recognition result] button to open the result editor (see use the result editor)

Submit a recognition task

To submit a recognition task in the Builder:

  1. In the domain's builder, click the Task list menu.
  2. Click [Submit recognition task] button.
  3. Select the tab for how you want to upload the input file.
    • Select from Object Storage tab: Select a file from the domain's input file path or its subdirectories.
    • [File upload] tab: Select a file from your local PC.
  4. Enter the task details.
    • Result file location: Select the Object Storage path where the result file will be saved.
      • You can select the domain's predefined result path and its subdirectories.
    • Language settings: Choose the language for recognition
    • Speaker separation: Select whether to use speaker separation.
    • Event detection: Select whether to detect events such as clapping, laughter, or music.
    • Keyword boosting: Select whether to use keyword boosting (for details, see Keyword boosting).
  5. Upload the input file.
    • When using Select from [Object Storage]: Select a file in the file selection area.
      • Double-click a folder to view its contents.
    • When using File upload tab: Drag and drop a file or click here. Drag a file into the area, or drag and drop a file, or click here to select one.
  6. Click [Start request] button.
    • When the task is complete, you can view the result JSON file in the result file path.
Note

A maximum of 10 speakers can be recognized in a single audio file.
When using the re-recognition feature, you do not need to submit the same audio file again.

Use the result editor

After a recognition task is completed, you can open the result editor to view the media file and recognized text together. You can also edit the recognition results and export them in the format you want.

Result editing interface

When you open the result editor, the Result editing interface appears. Here is an overview of its components:
clovaspeech-long-demo_ko

Component Description
① Export Edits and exports the recognition results.
② Input file
  • Plays or pauses the input file.
  • Displays the input file and recognized text information.
③ Filter Filters recognition results by speaker or confidence.
④ Speech recognition results
  • Displays recognized text in a timeline format.
  • Allows editing of detected speakers and text.
⑤ Event detection results
  • Displays detected events (clapping, laughter, music) in a timeline format.
  • Allows editing of detected events.
⑥ Playback Plays the input file and adjusts the volume.

Edit and export recognition results

You can edit recognition results and export them in the format you want using the result editor.

Note

You can export multiple result files at once using recognition result export.

  1. In the domain's builder, click the Task list menu.
  2. In the task list, click [Edit recognition result] button for a completed task.
  3. In the result editing interface, review the recognition results.
    • When playing the media file, the corresponding text in the timeline is highlighted.
    • [Speaker] button: Filter recognized text by speaker.
    • [Confidence in recognition results] button: Filter recognized text by confidence scores.
  4. Edit the recognition results if needed.
    • Edit speaker: Click the dropdown menu and select a speaker.
      • i-clovaspeech-speakerdelete: Delete a speaker.
      • i-clovaspeech-speakeredit: Edit a speaker's name.
      • [Add speaker] button: Add a new speaker.
    • Edit text: Click the marked row i-clovaspeech-resultedit to edit text, then click i-clovaspeech-resultedit2 to save.
      • To cancel edits, click i-clovaspeech-editcancel.
      • Edited content is highlighted in green.
  5. To export results, click [Export] button, enter the settings, and click [OK] button.
    • File name: Enter a name for the output file.
    • File format: Click dropdown manu and select the file format (json, smi, srt, txt, csv, xls)
    • Shift timeline: Push the timeline backward by a specified amount.
    • Export method: Select Object Storage or Download.
      • For Object Storage, click the file path to select the save location.
      • For Download, the file is saved to your local PC.

Export recognition results

You can export the result files of multiple tasks at once.

Note

Only Task Status with Completed task can be exported.

  1. In the domain's builder, click the Task list menu.
  2. Select the tasks you want to export.
    • Multiple tasks can be selected at once.
  3. Click [Export] button.
  4. When the export popup appears, review the settings and click [OK] button.
    • i-clovaspeech-edit: Edit the file name.
    • i-clovaspeech-delete: Remove a task from the export list.
    • File format: Click dropdown manu and select the file format (json, smi, srt, txt, csv, xls)
    • Export method: Select Object Storage or Download.
      • If Object Storage is selected, click the result file location to choose the path where the recognition result file will be saved.
      • For Download, the file is saved to your local PC.

Delete recognition tasks

To delete registered recognition tasks:

  1. In the domain's builder, click the Task list menu.
  2. Select the tasks you want to delete.
    • Multiple tasks can be selected at once.
  3. Click [Delete] button.
  4. When the delete task popup appears, click [Delete] button.

Keyword Boosting

Keyword boosting allows you to increase recognition accuracy for specific words. In the Keyword Boosting menu, you can register and manage keywords.

Keyword Boosting Screen

The basic layout of the Keyword Boosting menu is as follows:
clovaspeech-builder_keyword_ko

Component Description
① Menu name Shows the current menu name.
② Keyword management features Upload keyword files, download registered keywords, delete keywords.
③ Search Search keywords.
④ Keyword list List of registered keywords.

Add keywords

To add keywords you want to boost:

Note
  • You can add up to 1,000 words, and Korean and English are supported.
  • It is recommended to register 500 words or fewer. Registering more than 500 words may degrade performance.
  • Higher weight values increase recognition of similar words, which may affect accuracy.

You can add keywords using either method below:

Upload keyword file

To upload an Excel file and add multiple keywords at once:

  1. Prepare an Excel file containing the keywords and weight values.
    • Weight values must be integers between 1 and 5.
    • In step 4, click Upload template to view the keyword file template.
  2. In the domain's Builder, click the Keyword Boosting menu.
  3. Click [Upload] button.
  4. When the upload popup appears, drag a file into the window or click the Click here. Drag the keyword file from your local PC into the area, drag and drop the file with your mouse, or click here. Click the area to select the keyword file.
  5. Click [OK] button.
    • Overwrite: Enable the checkbox to delete all existing keywords and register only the new keywords.
Note

If the file contains keywords that duplicate existing messages, the upload will fail. Remove duplicate keywords or enable overwrite when uploading.

Add Keywords Individually

To add keywords directly in the Builder:

  1. In the domain's Builder, click the Keyword Boosting menu.
  2. Enter the keyword you want to add. Click the input area and type the keyword.
  3. Select a weight value from the dropdown menu.
  4. Click [Add] button.

Edit Keywords

To edit a registered keyword:

  1. In the domain's Builder, click the Keyword Boosting menu.
  2. In the keyword list, click the edit icon i-clovaspeech-edit for the keyword you want to modify.
  3. Click the keyword input field, edit the keyword, and then click i-clovaspeech-keywordeditcheck to save.
    • To cancel editing, click i-clovaspeech-keywordeditcancel.
    • To modify the weight, select a new value from the dropdown menu.

Download Keywords

To download registered keywords:

  1. In the domain's Builder, click the Keyword Boosting menu.
  2. In the keyword list, select the checkboxes for the keywords you want to download.
    • Multiple keywords can be selected at once.
  3. Click [Download] button.
    • An Excel file containing the keywords and weight information will be downloaded.

Delete Keywords

To delete registered keywords:

  1. In the domain's Builder, click the Keyword Boosting menu.
  2. In the keyword list, select the checkboxes for the keywords you want to delete.
    • Multiple keywords can be selected at once.
    • You can also delete an individual keyword immediately by clicking [Delete] button.
  3. Click [Delete Keywords] button.
  4. When the confirmation popup appears, click [OK] button.

Check API Call Information and Domain Information

In the Settings menu, you can check the integration information required for API calls and the information of the domain connected to the Builder.

Check API call information

To check the integration information required for API calls:

Note

You can call the API to send audio files for recognition and receive recognition results. For more information about calling the API, see CLOVA Speech API Guide.

  1. In the domain's Builder, click the Settings menu.
  2. Click the [Integration information] tab.
  3. Check the domain's Secret Key and CLOVA Speech Invoke URL.
    • [Create] button: Click to create or regenerate the Secret Key.
    • [Copy] button: Click to copy the information to the clipboard.

Check Domain Information

To check the information of the domain connected to the Builder:

  1. In the domain's Builder, click the Settings menu.
  2. Click the [Domain information] tab.
  3. Check the domain information connected to the Builder.
    • Domain information
      • Domain name: Name of the domain
      • Domain code: Code of the domain
      • Type: Service plan applied to the domain
    • Storage settings
      • Result file storage path: Object Storage path for saving recognition result files
      • Input file storage path: Object Storage path for selecting input files
      • [Go to Object Storage] button: Click to move to the Object Storage console interface