Skip to main content

Chat with Documents


info

Please see the Introduction for a list of requirements you'll need before you can chat with your data.

Path to your Chat-enabled Dataset

Fluree Cloud is built to be your easiest path to working with your data. You have a few options for layering a Fluree Data Model into your dataset. Choose the path that matches your source data and follow the guide with the goal of defining a data model that's compatible with the Fluree Data Agent.

  1. Upload Documents (this guide)
    You have unstructured documents like PDFs, Word docs, or text files.
  2. Upload Tables
    You have structured data from a relational database export or in a spreadsheet or set of CSV files.

Path: Upload Documents

Choose this option if you have text documents in one of the following file formats:

  • Plain Text or Markdown (.txt, .md)
  • Word (.doc, .docx)
  • PDF (.pdf)
  • HTML (.htm, .html).

At a high level, Fluree Cloud uses custom ensembles of machine learning models to extract entities and relationships against a combination of widely-used general ontologies we refer to as the General Knowledge Model. This method is best at analyzing files concerned with people, places, and things and the relationships among them. Think Wikipedia articles, news articles, or academic papers. Files that contain images, tables, or other non-textual data won't produce the best results with our current ML models.

Simply upload your text documents and our suite of automated data services will generate a localized "slice" of the General Knowledge Model that is specific to your data (this will become your Fluree Data Model) which will be used to extract structured data for you to transact to your dataset.

Here's the step-by-step.

  1. Navigate to your new dataset
    If you've just created your new dataset, you're already on your Dataset page. If not, navigate to your Datasets screen and click on your new, empty dataset.

  2. Call up the Files tab
    Click on the Files entry in the dataset menu on the left of the screen. This will open the Files tab for your dataset.

Dataset View indicating the Files tab
  1. Upload your text files
    Either drag and drop one or more text files onto the window or click on the Select Files button in the top right of the Files tab. Only the following file extensions will be accepted at this time: .txt, .md, .doc, .docx, .pdf, .htm, .html.

  2. Process your text files
    After your file has been uploaded, the Action button on the file's record will change to Process. Click this button to kick off the automated extraction process. Depending on the size of this file, this process may take some time. Feel free to navigate away from this screen and return later to check on the progress.

Files tab indicating the Process button
info

The screenshot above also includes a light-blue arrow that indicates a switch you can use to automatically process and transact the data from uploaded files without having to manually take action to move the process along.

  1. View and transact your extracted data
    When the processing of your file is complete, the Status column will indicate "Processed" and the Action button will change to Transact. You can click on the vertical ellipsis icon to the right of the Transact button to expose a submenu where you can Download JSON to see your extracted data. If everything looks good, click the Transact button to add the extracted data to your dataset.
Files tab indicating the Transact button and File Submenu
  1. Browse your data in View Data and chat with your data in Chat
    After you've transacted your data, you can navigate to the View Data tab to see your data in a tabular format by clicking on the View Data button on the file record or by clicking on the View Data menu item in the left-hand dataset menu. We recommend you browse the data in your dataset to get an idea of what concepts, terms, and relationships were extracted from your text files. This is the context the Data Agent will have when you chat with your data. When you're ready to chat with your data, click on the plus button on the Chat menu item in the left-hand dataset menu to create a new chat.
Files tab indicating options for viewing data and opening a chat session
Chat tab showing a sample question and answer with the Fluree Data Agent