Articles

How to Fine-Tune Handle?

Creating custom responses for your chatbot on the Handle dashboard is straightforward. The way your chatbot interacts with your customers largely depends on the content you provide and its structure. This process of training and setting up the chatbot is termed the Fine-Tuning phase. Essentially, it's about teaching the chatbot to provide specific information and communicate in your typical style.

How to Clean Your Data for Better Chatbot Recognition

If you're experiencing issues with Handle not answering questions thoroughly after uploading your files, or pasting information in the plain text section, it's likely due to unstructured data within your files or information. Unstructured data can include images, tables, special characters, and extra spaces that may confuse the chatbot. To resolve this, follow our simple data cleaning process outlined below:

Choose the Right Data Source Mode for Handle

The Data Source section in the Handle dashboard is where you can upload your training materials. Data Source Mode dictates how you'd like Handle to engage with your customers upon receiving a question. Handle provides three Data Source Modes to select from: General Data, Your Own Data, and Combined Data.

Ensuring Handle Understands Webpage Content

This article outlines a three-step process to enhance the performance of Handle when fine-tuning with webpage content. First, we address the potential issues when Handle scans webpages. Next, we offer a practical solution involving the manual transfer of webpage content to a document and uploading it to Handle for better understanding and response accuracy. Finally, we discuss testing Handle's comprehension post-upload and provide information on seeking further support if needed.

Improving AI Recognition for Detailed Lists or Characteristics

When using Handle to provide detailed lists and characteristics associated with specific subjects or topics, it's crucial to ensure that these lists are correctly connected to the right subject. Here's how you can enhance the chatbot's recognition and answer generation process.

How can I improve recognition in other languages?

Handle can understand and generate content in over 90 languages, and any fine-tuning information is adaptable to the language preferences of you and your customers. To improve recognition and content generation in specific languages, fine-tuning Handle with language-specific data is recommended, ensuring an enhanced user experience by addressing language nuances.

Teach Handle to Use First Person, Instead of Third Person, When Answering Questions.

Your Handle’s style is a reflection of the data you've provided for its training. If this data predominantly uses third-person references for yourself or your company, the chatbot might mirror this approach. As a result, it might reply in the third person when addressing customer inquiries, even when first-person responses would be more appropriate.

Set up Greeting Message and Interactive Options

Initiating a conversation with Handle, users are first greeted with a message. This initial interaction comprises two default components: the greeting message and the interactive options. In this article, we'll guide you through setting up and customizing your greeting message and interactive options to provide a more engaging experience for your users.