Data Entry Helpers

Suggested Entities (As Is)

Scenario 1. Active Lookahead - Suggestions Based on Reviewer Entries

sequenceDiagram
    actor Reviewer
    participant App
    participant Canopy AI
    Reviewer->>App: Enters New Entity Relationship
    Canopy AI->>App: Actively searches across all documents for all combinations of name (first name or last name) and any other element and attaches a suggestion to the document.
    Reviewer->> App: Validates suggestion
    App->> App: Automatically replicates entries to duplicate documents

Pros	Cons
We can sell the idea of saving time in review.	Makes suggestions for every single entry.
	Only works on incomplete Elastic search.
	Requires user to validate every entry.
	Deafeningly noisy to the point it is unusable.
	Makes duplicate suggestions

Scenario 2. Co-reference Suggestions - Suggestions Based on Learning Model

sequenceDiagram
    actor Reviewer 
    participant App
    participant Canopy AI
    Canopy AI->>App: Suggests relating entities to a name based on co-reference model
    Reviewer->>App: Validates suggestion

Pros	Cons
It is machine learning based and will get better with time.	Incredibly Inaccurate
Works without data entry.	Adds to the noise of active lookahead.
	Sometimes relation is besides the point.
	Customer still needs to validate the suggestion.

Entity Propagation Ideas

Scenario 3. Auto Add By Entity Group Rank

Phase 1 - Detected Entities

sequenceDiagram
    actor Review Manager
    actor Reviewer
    participant App
    participant Canopy AI
    Review Manager ->>App: Configures App to Suggest (one-by-one) / Automatically Add (bulk) Ranked Entities
    Reviewer->>App: Enters New Entity Relationship (doesn't require name)
    Canopy AI->>App: Rank Entity Suggestions based on Detected PII (refer to ranking algorithm)
    Canopy AI->>App: Consolidate on dedupable keys only on without conflict.
    Canopy AI->>App: Suggest Entities / Automatically Add Ranked Entities
    Reviewer->> App: Validates suggestion either one-by-one (Suggested) / By Batch (Auto Add)

Pros	Cons
Automatically adds raw entities	Requires some level of unsupervised entity consolidation
Adds entities based on dictionary based PII Detection	Doesn’t cover elements that have not been added by the user
Handles non-primary key elements	Overlapping Raw Entity Entries
Cuts down on entry time.	Won’t work for non-detected entities
Less resource intensive for suggesting
Less duplicate suggestions
Reduces number of entity entries

Can co-reference work better if used on top of the rank information? This way it covers elements added by the user.

Canopy AI-»App: Determine Suggestion to Update an Existing Entry or to Add a New One (Can we do this?)
What happens if I change what I edited a particular entry?
What happens if we delete the entry?

Phase 2 - Non Detected Entities

sequenceDiagram
    actor Dataminer
    participant App
    participant Canopy AI
    Canopy AI->>App: Flag entities that were non detected entities
    App ->> Dataminer: Alerts Data Miner that rules need to be expanded
    Dataminer ->> App: Adjust rule for PII element to expand detection.
    Dataminer ->> App: Rerun PII Detection for PII element only.
    Canopy AI ->> App: Either adjusts suggestions / automatically adds new entities

Phase 3 - Seed Review with Known Entity Relationships

sequenceDiagram
    actor Review Manager
    participant App
    participant Canopy AI
    Review Manager->>App: Map known entities which are not part of the compromised dataset (customers, employees, students, customers, patients) 
    Review Manager-->App: or add Canopy suitcase from previous project.
    Canopy AI->>App: Sets up auto suggestions or automatically adds the entities known entities prior to the review
    App->>Review Manager: Reports on the number of PII detection that have been entered vs that hasn't been entered.

Phase 4 - Suitcase Encrypted Existing PII

sequenceDiagram
    actor Review Manager
    participant App
    participant Canopy AI
    Review Manager->>App: Marks project as complete
    Review Manager->>App: Requests to suitcase project PII
    Canopy AI->>App: Creates Suitcase
    Review Manager->>App: Download Suitcase
    Canopy AI ->> App: Deleted Suitcase but maintains private key

Scenario 4. Auto Add By Element

sequenceDiagram
    actor Reviewer
    participant App
    participant Canopy AI
    Reviewer->>App: Enters New Entity Relationship
    Canopy AI->>App: Actively searches across all documents and adds an entity for each dedupable element it finds on a document. 
    App->> App: Automatically replicates entries to duplicate documents

Auto Add
Dedupable Element

Pros	Cons
Automatically adds raw entities	Customer rejected several variations of this solution.
	Raw entities are added one-by-one shifting burden to consolidation
	Creates duplicate entries
	Cannot handle non primary keys
	Causes confusion for reviewers because they may want to combine entities auto added on a page or update entity anyway
	Detection based on elastic and not dictionary.