Tool to semi-automate training data creation for AI “SMART ANNOTATOR” launched


Studio Ousia announced the launch of SMART ANNOTATOR, a tool that uses our natural language processing technology to semi-automate the process of generating training data for natural language processing (NLP) AIs. To build a practical natural language processing AI, a large set of training data needs to be compiled for the AI which takes a lot of manpower and has proven to be a hurdle in the adoption of NLP-AIs. By using SMART ANNOTATOR, the training data generation process can be semi-automated, greatly reducing this cost.

SMART ANNOTATOR is a web service that uses advanced NLP-AI to automatically cluster (grouping of texts that are similar in meaning) large amounts of text data. It can be used to generate training data for a wide variety of AIs, such as those used in chatbots, text classification of news or tweets, and analyzing inquiries and complaints. SMART ANNOTATOR can also be used for text-clustering of large data for visualization purposes.

Studio Ousia provides the high performing question-answering system “QA ENGINE”. SMART ANNOTATOR was developed to meet the needs of chatbot and customer support managers that had large amounts of text data but came to Studio Ousia seeking support in turning the raw data into learning datasets for their AIs.



  1. Instant clustering of uploaded text data using AI
  2. Easy association of generated text-cluster and uploaded class information

Use cases for SMART ANNOTATOR:

  1. For chatbot administrators
    • Generation of training data for chatbot installment or maintenance
    • Analysis of queries to update FAQ or marketing use
  2. For marketing managers
    • Inquiry and Voice of Customer (VoC) Analysis
    • Analysis of free answers from questionnaires
  3. Text classification
    • Classification of news and tweets
    • Claim detection

Service format

SMART ANNOTATOR is a monthly subscription service. Subscription starts at 500 USD/month for 1 user ID. No initial fee is required.


Please contact us at qaengine@ousia.jp