Perfecting Cutting Edge Large Language Models

Text Generation

Mindy Support played a pivotal role in facilitating the creation of meaningful dialogues for our client, culminating in a valuable dataset earmarked for the refinement and training of their AI chatbot. The collaborative efforts with Mindy Support resulted in the development of authentic conversations, strategically curated to enhance the cognitive capabilities of the client’s AI-driven chatbot, thereby ensuring a more sophisticated and responsive user experience.

The client develops Conversational AI for big customers and requested our services for text content generation. They requested the creation of 5000 scenarios resembling Fintech/Consumer Banking support conversations (English) between two individuals (each comprising approximately 12 sentences). Subsequently, we offered semantic segmentation services utilizing this dataset

In addition to the sheer volume of the dataset, the client needed the participants to discuss various topics relating to customer support and sales. Even though a lot of people and texts would need to be generated, the client still had a tight deadline of 2 months to complete this project. This is why they turned to a well-known data annotation company, like Mindy Support, for assistance.

Mindy Support has had a longstanding relationship with the client, dating back many years. Over the course of our partnership, Mindy Support has consistently proven itself to be a trusted provider of a wide range of data annotation services that could help meet tight deadlines and scale projects without having to sacrifice quality.

We started off by creating a plan on how we will recruit the needed 500 people to participate in the project. Since we are one of the largest data annotation providers in Europe, we already have recruiting processes in place, which include avenues like job sites, referral programs, social media PR, and many other tools. Once our team was in place, we proceeded to the onboarding stage, where we conducted training for all of our participants. When everybody understood the client’s requirements, everything went smoothly, since all of our team members had extensive experience with both text generation and data annotation. 

Thanks to our combined efforts, we were able to meet the deadline set by the client and achieve a quality score of 98%+ for the generated texts. The client was very pleased by both the pace of the text generation work and the quality and decided to further expand the project to include additional languages such as German, French, Italian and Turkish.

• 500 participants recruited 
• 98%+ quality score on all generated texts
• Met the client’s requirements to complete all the work within 2 months