Entity Extraction

Entity Extraction allows you to extract important data sets from various types of documents. For example, you can extract the signing date and the contractors’ names from a contract, or shipping status and product name from an invoice.
For most, you simply upload the documents and hit a button. Alli’s AI will take care of the rest!

Adding Entity Set

First, find the Entity Extraction icon on the left navigation bar in your dashboard, then click the ‘+ Add Entity Set’ button.
From here, you can define the parameters for the entity set. You also have the ability to choose how often the extraction will run. Below is an example.
  1. Enter the name of this entity set.
  2. Enter a description for this entity set.
  3. Select when to run the entity extraction. You can select from ‘Only when the Extract button is clicked’, ‘When the hashtags are added to the documents’, or ‘Every midnight’.
  4. Define the document type here.
  5. Cancel or submit the entity set.
If “Paragraph Style” document type is chosen, you must define the entities. Put the name of the entity and the question to find it. For example, you can name an entity as ‘Shipping date’, and put a question of ‘When did it ship?’.
Once you submit an entity set, you can see the list of registered entity sets in the Entity Extraction menu.

Running Entity Extraction

Once you add one or more Entity sets, you can manage the sets and extract the entities from the entity set list.
  1. Click to add more Entity Sets.
  2. Search entity set by name.
  3. The last entity extraction result will be displayed here.
  4. Click to run the entity extraction right away.
  5. Click to duplicate the Entity Set.
  6. Click to delete the Entity Set.
If you click anywhere on an entity set, you can see the details of the entity set under the General tab.
  1. Click to go back to the Entity Set list.
  2. Details of the Entity Set.
  3. Click to run the entity extraction right away.
  4. Upload more documents to the Entity Set.
  5. Manage the Extraction Model training and versioning.
  6. Internal description of the Entity Set.
  7. Extraction frequency information.
  8. The last entity extraction result will be displayed here.
  9. Edit the details of the Entity Set such as it’s name and a Description.
  10. View files within the Entity Set as well as the extraction results.
  11. Preview and edit the entities to be extracted from that document.

Preview the Entities to be Extracted

  1. Click to go back to the Entity Set list.
  2. Bulk select, filter, and bulk delete entity sets
  3. Hide the sidebar.
  4. Preview of the document including the KEY and VALUE pairs that are currently being extracted.
  5. Edit a KEY and VALUE pair.
  6. Delete a KEY and VALUE pair.
  7. Edit all KEY and VALUE pairs.
  8. Toggle on or off if document will be used for training.
Instead of searching for a specific KEY or VALUE, use filtering to easily navigate the extraction results.

Edit the Entities to be Extracted

  1. Filter entity sets by KEY and VALUE results
  2. Edit a KEY or VALUE. More than one data point can be selected.
  3. See all data points that can be extracted from the document.
  4. Reset the KEY and VALUE set.
  5. Delete the KEY and VALUE set.
  6. Cancel or save changes made.
  7. Toggle on or off if document will be used for training.
If a value has not been identified yet for a KEY or VALUE, we have added in the ability to leave it blank for the time being as shown below.

Adding a New Entity Group

As a new feature to Alli, we have added the ability to create a new entity group to be extracted. Simply scroll to the bottom of the list and click on Add Result Group.

Checking the Result

You can see the extraction results under the Result tab in each Entity set’s details.
  1. Click to go back to the Entity Set list.
  2. Entity extraction results will be displayed under the Result tab.
  3. Click to run the entity extraction right away.
  4. Upload more documents to the Entity Set.
  5. Retrain the Extraction Model.
  6. Search specific entity extraction result by keyword.
  7. Click to decide how to run the keyword search. You can select ‘Search from entity name’ or ‘Search from extracted result’.
  8. Click to convert between ‘Show latest result only’ and ‘Show all’. Default is Latest result only, so you can only see the extracted entities from the last run.
  9. The list of extracted entities and details.
  10. Identify which result is from which model.
  11. Click to see where exactly the result was extracted from the target document and edit if polishing is needed.
  12. Click to delete individual entity extraction result.

Managing the Model

Once you add one or more files to an Entity set, you can manage the training for that entity set after the results are checked.
  1. Click the check box next to the desired files for training.
  2. Click the three dot icon next to the trashcan icon.
  3. Select “Turn on training selected” to use the selected files for training the entity set.
  4. The toggle under “Training” can be used to turn on training for individual files.
To retrain the model, click on “Retrain Extraction Model”. There will be two options to choose from.
  1. Only train the AI model after making adjustments to the results.
  2. Train the AI model and re-extract all entities from the documents.
To manage the current model or deploy a different model, click on “Manage Models”
  1. Deploy an extraction AI model to compare
  2. Edit the name of the extraction model.
  3. Delete the extraction model.
When deploying a model, Alli gives you two options of deployment.
  1. Deploy the AI model only
  2. Deploy the AI model and re-extract all entities from the documents.
If you deal with a lot of forms and in need of extracting key entities from them, this new feature will greatly boost your productivity by taking care of the most tedious, repetitive tasks for you. Please reach out to your account manager or biz@allganize.ai if you have any feedback regarding Alli’s entity extraction feature.