How To Reduce NLP Model Biases For Better Text Analysis (with Dan Erickson)
Monday, May 23, 2022
In this episode of The Product Manager Podcast, I spoke with Dan Erickson, CEO and Founder of Viable, a company that helps businesses learn from customer feedback through advanced natural language processing.
Dan brought unique insights from his experience developing NLP tools that aggregate customer feedback across multiple channels to identify themes and produce actionable insights. At the time of our conversation, Viable was using a combination of in-house and third-party models, including GPT-3, to analyze qualitative data.
We explored:
- The fundamentals of Natural Language Processing (NLP) and how it’s used in product development
- How bias emerges in language models and practical ways to identify and reduce it
- The challenges of analyzing qualitative data at scale and turning it into actionable insights
- How product teams can use NLP tools ethically and effectively
This conversation highlighted both the power and potential pitfalls of using AI-driven language tools in product development, with practical advice for product managers looking to leverage this technology responsibly.
This episode was part of my work as founding host of The Product Manager Podcast (2021-2022), which has since grown to reach over 100,000 downloads.