Technology Stack
Natural Language Processing
Challenge
The client had a lot of manual work assessing qualitative open text responses on surveys from their customers in order to create reports. They were after an automated way that reduced the human error, biases and that stood up to their quality standards to produce these reports. The reports combined both these qualitative answers with the quantitative demographic survey responses. Â
Solution
Blackbook built a solution using two highly regarded open-source LLM’s to solve the clients challenge. The first model was able to ingest all the responses from a question to identify the key topics, then assign the individual responses to topics with a confidence score. The second model assigned a sentiment score to the responses, as well as pulling out key phrases and entities so the client could understand further insights. This data, along with the quantitive demographic survey data was modelled into a dashboard with automated reporting so the process was an end-to-end solution.
Outcome
The client saved significant time and resources by being able to automate a main part of their business processes. Analysis that usually took several resources weeks, now took under a day to produce and allowed the resources to spend time doing higher intelligence work looking for deeper insights into the data. Â