LLM Support in Documents

Updated on May 10, 2024

We firmly believe that our customers should have the ability to automate tasks by leveraging their own documents with the assistance of a large language model. Our capabilities now encompass Q&A generation, document summarization, and custom prompting tailored specifically for businesses. By tapping into knowledge that may elude a general-knowledge trained LLM, we ensure a level of expertise and precision that is truly invaluable in a professional setting.

These features will only be available for an Alli Answer Generative or LLM enabled project. For document management, please see No. 1 of the Table of Contents here, however you will note that the dashboard does have a different appearance and functionalities change as it is a guide for our traditional Alli Answer product. To begin, navigate to Documents within the Knowledge Base.

  1. Document title search to find a specific document
  2. Preview of the document
  3. Q&A Generation from document
  4. Summarization of document
  5. Custom prompting of document
  6. Upload documents

LLM Models #

These models are the current offerings for what can be used for LLM support in documents out of the box. We are not limited to these three models – however have found that they provide the best responses.

Text-Davinci-003 – well-suited for generating text on a wide range of topics and writing styles. It possesses extensive knowledge and supports interactive responses. It can be used in the following scenarios:

  • When answering questions that require detailed information or specialized knowledge
  • When engaging in conversations with users to explore information or generate documents

GPT-3.5-Turbo – characterized by its fast response times and performance optimization. With enhanced accuracy, it enables high-quality text generation. It is suitable for the following scenarios:

  • When real-time interaction or response is required
  • When more efficient resource usage is desired

GPT-4 – with its larger model size, improved responsiveness, and advancements in deep learning, is expected to excel in the following scenarios (although it has not been officially released yet):

  • When complex text generation or advanced semantic understanding is required
  • When leveraging the latest natural language processing capabilities is desired

Q&A Generation #

Once documents have been uploaded, you have the ability to generate Q&A pairs. This is helpful for both internal and external use cases where FAQ’s are needed based on given documentation. This is very helpful for generating content for internal employees who need HR questions answered, or to customer support needing to foresee what questions may be asked.

To begin, click Generate

  1. Choose which LLM model you would like to use. Out of the box, we offer Text-Davinci-003, GPT-3.5-Turbo, and GPT-4 all from OpenAI.
  2. See the prompts that are being used
  3. View and edit the prompt from the Prompt Management page
  4. See the full list of prompts used for Q&A Generation
  5. Select a page range if desired from the document. If the first two pages are only needed, select “1 To 2” for a page range. Leave blank if the entire document page range is desired.
  6. Create and view a sample of Q&A questions before generating for the entire document
  7. Cancel or Generate a response.

The generated Q&A pairs are registered within the Q&A Candidates list. Like all Q&A candidates – you can remove, edit and, tag your candidates before deploying it live.

Summarization #

If a document summarization is required either for internal use or sharing with outside sources, this functionality is very helpful for generating content. The summaries can be a typical summary, chronological, or a conceptual summary based on the prompt chosen or made.

To begin, click Generate

  1. Choose which LLM model you would like to use. Out of the box, we offer Text-Davinci-003, GPT-3.5-Turbo, and GPT-4 all from OpenAI.
  2. Choose the prompt you would like to use. We will discuss prompt management later in the document.
  3. View the contents of the prompt and double check it is the right choice.
  4. Select a page range if desired from the document. If the first two pages are only needed for a summarization, select “1 To 2” for a page range. Leave blank if the entire document page range is desired.
  5. Cancel or Generate a response.

Utilizing a Custom Prompt #

In order to create a custom prompt, we must utilize our prompt manager within the settings page. To learn more and create a prompt, please see our guide here for prompt management.

Once we have our new prompt created, let’s use it for our documents. To begin, navigate to the documents and custom prompt tab.

  1. Choose which LLM model you would like to use. Out of the box, we offer Text-Davinci-003, GPT-3.5-Turbo, and GPT-4 all from OpenAI.
  2. Choose the prompt you would like to use.
  3. View the contents of the prompt and double check it is the right choice.
  4. Select a page range if desired from the document. If the first two pages are only needed, select “1 To 2” for a page range. Leave blank if the entire document page range is desired.
  5. Cancel or Generate a response.
Table of contents