Categorise Customer Feedback from Online Forms with Cohere

Log in to your Cohere account

Begin by opening your web browser and navigating to the Cohere website. On the homepage, you'll usually see a "Log in" button, typically positioned in the top-right corner of the screen. Click this button, which will take you to a login page. Here, carefully enter your registered email address and password into the respective fields. If you haven't created an account yet, look for a "Sign up" or "Get Started" option nearby to guide you through the registration process. You'll know it worked when you see your Cohere dashboard or main account page, which typically displays an overview of available AI tools and services.

Navigate to the Playground

The Cohere Playground is a web interface where you can interact directly with their AI models without needing to write any code. From your Cohere dashboard, look for a navigation menu. This menu is usually found on the left-hand side of the screen, or sometimes along the top. Scan the options and click on the one labelled "Playground". This action will open a new interface specifically designed for experimenting with different AI models. If it looks different, scan the navigation menu for an option like "Try it out" or "Experiment", or a section dedicated to "Models", which should lead you to a similar interactive environment. You'll know it worked when you see a large text input area, often with options to select or configure different AI models, indicating you're in the right place to begin your tests.

Select a text generation model

Cohere offers several AI models, each designed for different tasks. For classifying feedback, we need a model that excels at understanding instructions and generating text. In the Playground, you'll typically find a section for model selection, often a dropdown menu labelled "Model" or "Base Model". Look for and select a model such as "Command" or "Command R+". While model names can change, these are Cohere's general-purpose models known for following instructions well. If it looks different, look for a dropdown labelled "Model" or "Base Model" and select any option clearly intended for general text generation or conversational AI, rather than models for embeddings or reranking. You'll know it worked when the Playground interface updates, showing input fields and controls (like 'temperature' or 'max tokens') specific to text generation tasks, confirming your model choice.

Input your feedback and categories as a prompt

Now you'll provide the AI with clear instructions, known as a 'prompt', detailing what to do. The prompt will contain your customer feedback and a list of possible categories. In the main text input area of the Playground, type your instruction. Start by presenting the customer feedback, then clearly list the categories you want the AI to choose from. Make sure your categories are distinct to help the AI make an accurate choice.

You'll know it worked when your complete feedback and category list are clearly visible in the prompt box, ready for the AI to process.

Run the model and review the output

With your prompt ready and the appropriate model selected, it's time to send your instructions to the Cohere AI and see its classification. Look for a "Run", "Generate", or "Submit" button, typically located below the prompt input area or on the right-hand side of the screen. Click this button to send your prompt to the AI. The AI will then process your feedback based on the categories you provided, and a new section will appear, displaying its response, which should be the identified category for your feedback. You'll know it worked when Cohere displays the AI's generated response in an output box, and this response is one of the categories you listed in your prompt.

Refine your categories for better accuracy

If the AI's classification isn't quite right or doesn't match your expectations, you can adjust your categories or the prompt to improve the results. This iterative process is a common and important part of working with AI. Consider why the AI might have chosen an incorrect category: perhaps your categories were too broad, they overlapped, or the feedback itself was too vague. Go back to your prompt in the Playground and try rephrasing your categories to be more distinct or add more specific context. Then, click "Run" again to see the updated classification.

You'll know it worked when the AI consistently categorises your feedback into the specific types you intended, leading to clearer and more reliable analysis.