Updated on 6/2/2021

  • You can now update the search result footer for each document.

Machine Reading Comprehension (MRC) can be used to answer questions from large corpus of documents with high accuracy. This is particularly useful for customer support and auto-generating Q&A knowledge base to streamline communication.

In Alli, you can upload your customer support manual, return policy, etc. to be processed by our MRC technology which will automatically extract answers from documents you uploaded when the user asks a question.

Uploading Documents

To start, go to the Documents tab on the Knowledge Base page and upload a document. We currently support txt, xlsx, docx, doc, ppt, pptx, jpg, jpeg, png, and pdf format.

You can easily organize and manage the uploaded documents using hashtags. Click on the edit icon to add hashtags to the documents or update the footer for search results from this documents.

As default, Search result footer has a link to download the source document. You can add any text, links, images, videos, or files using the editor.

The search result footer will be shown at the end of the search result extracted from the document.

Finding Answers from Documents

Using the Answer from Documents node, you can let the customers find answers from uploaded documents in the conversation with Alli. Or when a user tries to find an answer from your Q&A database and there isn't a relevant Q&A, Alli AI automatically searches for answers from your uploaded documents.

You can find Alli AI at work in the Candidates tab on the Knowledge Base page:

There are three main states to Alli's AI:

  • AI is working on the answer.
  • AI failed to find an answer.
  • The answer is automatically filled in if a match is found. However, this option is not live or active by default. You need to add it to Q&A by clicking the '+' icon at the bottom right.

Once added, the question and answer pair is active and ready for your users. You can find all the active Q&As in the Q&A tab:

Learn More About Allganize's Technology