IBM Watson自然語言分類器正式上線

News from: iThome & IBM Cloud Blog

Watson自然語言分類器整合了IBM的深度學習服務（Deep Learning as a Service），一次最多可以匯入2萬行的資料，作為訓練素材。

近日IBM宣布，自家AI平臺Watson自然語言分類器（Natural Language Classifier）正式上線，現在該工具已經內建於IBM Watson Studio，透過該開發平臺可以串接起資料科學開發應用流程，加速深度學習、機器學習應用的開發。IBM表示，該分類器與Watson Studio整合後，仍保有既有API功能，開發者可以在開發平臺中訓練、測試此自然語言分類器。

結合Watson自然語言分類器與機器學習，使用者可以建立、訓練客製化的分類器。IBM進一步解釋，該分類器整合自家的深度學習服務（Deep Learning as a Service），一次最多可以匯入2萬行的資料，作為訓練素材。而現在該分類器所支援的語言，包含英文、阿拉伯文、法文、德文、義大利文、日文、韓文、葡萄牙文，以及西班牙文。

目前IBM Watson Studio開發平臺所支援的深度學習框架，包含Keras、TensorFlow、Pytorch以及Caffe。此外，該平臺也支援熱門的資料科學、深度學習工具，例如Jupyter Notebooks、Anaconda、RStudio。

------------------------------------------------------------

Watson Natural Language Classifier

Independent of industry vertical or use case, natural language processing has become central to all user workflows. From developers to data scientists, an optimized workflow is essential.

“At the root of all natural language processing is text classification”

From classifying financial risk and compliance to categorizing service queries and messages, Watson Natural Language Classifier enables you to create, train, and evaluate custom classifiers through machine learning. Important features include the following:

• Classify natural language text: Classify strings of text into categories of your choice.
• Leverage IBM Deep Learning as a Service: Train your classifier with up to 20,000 rows of training data.
• Multi-lingual support: Support is available in English, Arabic, French, German, Italian, Japanese, Korean, Portuguese (Brazilian), and Spanish.
• Multi-phrase classification: Classify up to 30 separate text inputs in a single API request.

A frictionless user journey

Today we are happy to GA the Watson Natural Language Classifier tooling in Watson Studio. While maintaining existing API functionality, you can train and test your classifiers within Watson Studio.

When redesigning the tooling, it was important to keep the full end-user experience in mind. By addressing common pain points in the beta and old toolkit, the new tooling provides an improved user experience for training classifiers.

In one environment, you train, build, and test your classifiers. Additionally, with Watson Studio, you can import other services, notebooks, frameworks, and models into your projects.

Training a support ticket classifier

Note: Users of the old toolkit can have their data migrated to the new one by following these instructions.