A Survey to Understand Challenges of Deploying Text Classification

Hello everyone,

As more and more machine learning libraries are developed, it becomes much easier to build a text classifier. However, there are still a lot of challenges, ranging from collecting the training data, achieving high accuracy, making the classification fair for different groups of users, defending against malicious input, etc. As a group of researchers from MIT, we are curious about what are the challenges for industrial practitioners currently having to deploy text classifiers. If you have experience in deploying text classifiers, I wish you can spend 15-20 minutes filling out this survey to help us understand the challenges. You will also enter a lottery for a $25 gift card.

Link to the Survey

Why should you participate?
Have you ever encountered a problem when deploying a text classifier, and could not find a good solution? We believe that there are common problems in the deployment process, while some of them could have a better solution. Our research is to understand the actual challenges in the deployment of text classifiers, and to establish connections between these challenges and future academic research. We will summarize the results into a position paper to call on researchers’ attention to solving problems in the practical deployment.

Hey, nice survey. I just filled it out.

What I was missing, though, is more information about the researchers behind this survey, like
MIT, AI Lab XYZ and the names of a couple of people. I think adding this information to the survey would make it much more trustworthy and likely that people fill it out.

Also, when will you share the results of the survey?


Thank you for your quick response. I’m Lei Xu from MIT Data to AI Lab. This survey is part of my PhD research on deployable and robust text classification.

About the timeline, it highly depends on how many responses we can collect. We are targeting at summarizing the results into a research paper in a three-month timeline. I’ll keep you updated.