All Products
Search
Document Center

Manage custom models

Last Updated: Oct 18, 2019

Overview

Alibaba Cloud Intelligent Speech Interaction provides basic speech recognition models, including universal models and models specific to certain fields, for example, the 8 kHz telephone customer service and quality inspection model and the e-commerce model. If you have accumulated a large amount of historical data in your field, you can use the data to optimize your models.

On the speech recognition customization platform, you can create a model based on a basic model and upload a training corpus. After training the model based on its training corpus, you can effectively improve the speech recognition rate in specific scenarios. This method is especially suitable to optimize the recognition effect of proper nouns and high-frequency words.

Prepare a training corpus

Prepare a training corpus according to the calling limits and optimization suggestions. For more information, see the training corpus description in the introduction of custom models.

Activate the trial edition

To make you better experience the optimization effect of the customization platform, we offer you a one-month trial if you have not activated the customization platform.

ASR setting

Trial rules for the customization platform:

  • You can get a one-month trial if you have not activated the customization platform.
  • You can activate the business edition at any time during the trial period.
  • After the trial period expires, if the business edition is not activated, the custom model will be unpublished and become unavailable.

After activating the trial edition, you can use the customization platform. The remaining trial period will be prompted on the graphical user interface (GUI). You can activate the business edition at any time during this period.

Create a custom model

The console allows you to train custom models. Before creating a model, upgrade a speech recognition service to the business edition on the Speech Recognition tab of the Service and Purchase page, as shown in the following figure. For more information, see Activate services and purchase resource packages.

Upgrade to the business edition

After upgrading the speech recognition service to the business edition and purchasing models, choose My Resources > Custom Models in the left-side navigation pane, and then click Create Model, as shown in the following figure.

Create a model

Set Basic Model Type to a model type that is applicable to the corresponding scenario, for example, the e-commerce model or the 8 kHz telephone customer service and quality inspection model.

Pay attention to the restrictions on the training corpus. Upload a training corpus, and then click Save, as shown in the following figure. Then, you can use the corpus to train your models.

Configuration

1. Train a model

Not trained

Note: After a model is created, it is in the Untrained status. Click Training Model to train the model.

2. Wait until the training is completed

Train

Note: The model is in the Training status when it is being trained. After the training is completed, the status changes to Applied.

3. Apply a model

[Apply a model]

Note: When the model is in the Applied status, you can apply it to projects. On the All Projects page, click Project Settings on a project card. On the page that appears, click Switch Model, select one of your custom models, and then click Publish to apply the model to the project.

You can test the speech recognition effect of your custom model in real time in the test section.**

4. Delete a model

Take the recording file recognition service as an example. On the Service and Purchase page, click Alter self-training model tool in the Actions column corresponding to the recording file recognition service.

Delete a model

On the page that appears, select Close, and then click Activate in the lower-right corner.

Select Close