I want to triage answers answers from students based on prior answers to the same question and work out of they are good or bad.

Question

Hi all,

 

I'm working on an education solution which will reduce the amount of time needed for an examiner to conduct 121 interviews with students.  Typically these can take one to two hours.

 

The intention is to replace the time the examiner spends with the student initially and focus only on answers that the student did not answer well enough. 

 

I have a collection of existing interviews with written reports on how well the student did on each question for training.

 

I see the process as follows:-

 

• Pose a series of questions to a student in a linked video.
• Record their responses in video.
• Use Azure Video Indexer (or similar) to extract a transcript of the student's answers
• Use A.I. to reference the answers against previous answers that have additional information about the quality of response.  E.g. the student mentions the correct topics.
• Create a report that tells the examiner which questions the student did well on and which ones need attention.
• Link to video of each question.
• Examiner meets the student to only discuss the problem questions.

 

Any thoughts or suggestions would be appreciated.

 

Thanks

 

Richard

Answer 1

@Richard Wray

Thanks for reaching out to us. Your idea sounds promising and it's definitely feasible with current technology. Here are the steps on how you can achieve this with Microsoft Azure services, please take a look at it and let us know if you have any questions:

1. Recording: You can use any video recording tool to record the student's responses.
2. Video Indexing: Upload the video to Azure Video Indexer. It uses AI to extract insights from videos, such as a transcript of the student's answers.
3. Training a Model: Use Azure Machine Learning to train a model with your existing interviews and reports. The model should be trained to understand what constitutes a good response based on your training data.
4. Analyzing Answers: Once the model is trained, you can use it to analyze the transcriptions from the Video Indexer. The model should be able to evaluate the quality of the response, similar to how an examiner would.
5. Generating Reports: Based on the analysis, you can then generate a report using Azure Logic Apps or Azure Functions. The report can detail which questions the student did well on and which ones need attention.
6. Link to video: The report can also contain a link to the specific timestamp of each question in the video stored in Azure Blob Storage or Azure Media Services.
7. Reviewing: The examiner can then review the report and the linked videos to prepare for a follow-up meeting with the student. This process can help to significantly reduce the time examiners spend on initial interviews and allow them to focus on areas where the student needs improvement.

Since Azure OpenAI can work for step 3, 4 and 5, you can use GPT-4 model instead of the machine learning model to reduce your effort, please be aware that Azure OpenAI need access approval, but you can leverage other LLM in Azure Machine Learning Studio with no access approval.

I hope this helps, please have a try and feel free to post any questions in the forum.

Regards,

Yutong

-Please kindly accept the answer if you feel helpful to support the community, thanks a lot.