Success Case

NER for developing AI Tutors to Build a Smart Learning

NER automation

Byoungjoon Min
NER for developing AI Tutors to Build a Smart Learning

Overview | Grow Your Learning Platform with AI Tutor Development


As the EdTech industry market has expanded, so has consumer demand for personalized content. However, unlike the market's need for a variety of educational services and learning content models, the ability to actually customize them is not easy to find. We needed a platform to create our own content, as well as AI tutor development.

NER technology can be used to analyze large amounts of text, such as articles, research papers, and books, to identify relevant topics or concepts and keywords. This can help curate and organize training materials tailored to specific subject areas or learning objectives.

In order to create unique educational content, our client wanted to use NER to interpret the syntax of text and introduce a foundation for creating content that is customized for students rather than similar content.


Problem | From automating NER for AI tutors to managing workers


In order to secure its own educational content and services, Minigate needed to build large-scale data for AI training in fields such as science and humanities.

This project was unique in that it required platform customization. While they had the manpower to build the data, they needed a platform that could process the data into guided learning. Most of the functions for building the data needed to be further developed.


Implementing the ability to step through the same data and add new labels to different tasks was a top priority, and for some data, we needed to add the ability to enter the label values into a CSV and have the platform display the label values when uploaded.


In addition, the client requested a guide to utilize and operate the platform according to the content of Minigate's project. Based on the guide, we agreed to provide training to the client on project operation and management, such as worker settlement/management/statistics checking/data upload/assignment/adjustment.

Datahunt began a project to customize the SaaS platform to the client's wishes. In the end, we customized the features optimized for the client, and the collaboration began in the form of training the client's labelers.



Solution | Establishment of 3 stages of NER processing and training to utilize the platform for AI tutor development


The three steps of NER processing requested by Minigate to DataHunt are as follows.

  • Step 1. Recognize text objects
  • Step 2. Translate text into a foreign language
  • Step 3. Check the object recognition and translation process


Datahunt NER Smart learning autolabelling
Example of real work (auto labeling)

We've also designed it to minimize workflow and improve quality by providing features such as billing of workers involved in the project, progress management, and rejection rates.

Once the customization was complete, we stayed in constant communication with the client to get feedback and improvements on the finished platform. As much as customization is about functionality, it's also about the stability of the platform.

Even after the platform is built, we need to respond to queries and enhancement requests in real time. To reduce the number of queries, DataHunt provided additional non-face-to-face platform guide training.

데이터헌트 NER 작업 admin
The three stages of NER requested by the client consist of text object recognition, foreign language translation, and object recognition translation and verification.


Result | AI Tutor Development and NER Auto-labeling Platform


After a long project, the customization of the SaaS platform was finally completed.When a labeler enters data, it starts organizing text data for basic sciences/humanities, etc. and object recognition for words. After that, foreign language translation for text (sentences) is done. Finally, object recognition and foreign language translation are checked for accuracy.


Even after customizing the platform, we actively reflected the customer's improvement requests to improve the convenience and UX of the existing features. By listening to the voices of actual crowd funders and managers, we significantly increased the convenience and efficiency of operation and management.We were able to achieve great satisfaction not only for the client but also for the project staff.

Talk to Expert