About

An Analysis node in the Acodis platform is a workflow node that structures or extracts data from documents in a use-case specific and automated manner.

An analysis in essence consists of:

Analysis steps: modular processing steps that need to be performed on the documents to structure or extract the data. This could e.g. be a structure step find titles & title hierarchy or a data points step to tag & extract text.
Document annotations: that further specify the exact meaning of each annotation step among the analysis steps in the context of this use-case. E.g. for a structure step, the annotations can be used to define in the context of this use-case (a) what are the titles and (b) on what hierarchy level they appear in the table of contents.
An analysis model: that executes the analysis steps in order and automatically predicts the document annotations based on a specific set of documents

The analysis steps in the “SDS Extraction” analysis are “Native Text” and “SDS Structure” as seen on the left. The annotations for the SDS Structure step are applied on the document preview to further specify which titles are needed for this use-case and of what hierarchy level.

Settings and Options

In the workflow editor, when an Analysis Node is selected, the following configurations are available in the right-hand side panel:

Model Version: choose which model and analysis version should be used for the selected analysis node
Manage Models: navigate to the analysis model page to inspect available models or create new ones
Configure Confidence Warnings: select the confidence levels below which a prediction requires human validation

Analysis node - settings side panel

Confidence Warnings v38

The “Configure Confidence Warnings” button opens a configuration dialog that gives you fine-grained control over how model predictions are validated, allowing you to define minimum confidence thresholds
Predictions that fall below the specified thresholds will automatically trigger warnings and, depending on your automation settings, may require manual review and verification.

By setting confidence thresholds, you define how strict the system should be—lower thresholds allow more predictions through, while higher thresholds provide more control and accuracy by involving human review when needed.

Thresholds can be applied hierarchically, meaning parent steps can have one threshold, and nested data points can have their own stricter ones if needed.

When you select a specific item in the hierarchy, an interactive confidence chart appears on the right side of the dialog
This chart helps you select a threshold that balances automation and accuracy, using insights from the selected model’s evaluation on training data to provide a data-informed starting point.

In the confidence threshold configuration dialog of this analysis different thresholds are set. E.g. for the data point “Version” a threshold of 65% is set. Hence a document in production where a prediction for “Version” is made with a confidence of lower than 65% will produce a warning. In the review node it can be configured to require human validation on warnings.

Prerequisites

This feature is available to users in the role of ADMINistrators Workflow managers Case handlers

Export Assets

Analyses can produce export assets (also called export artifacts) which can then be downloaded or used in a Workflow . Export assets are typically a piece of structured information as defined in the analysis. For example it can be a digital twin of the document in the format of an XML or a JSON file containing the specified data points of a document.

Export assets can come from three main sources:

From the analysis itself, which are always available. The possible formats are
From a Data Points step. Refer to https://acodis.atlassian.net/wiki/x/AQAGNg
From a Custom Step

The export assets from the analysis itself can be directly downloaded using the analysis document viewer or exported as part of a workflow in any of the export channel nodes

Analysis Versions

An analysis can be versioned v24 . This means that all the configurations regarding the analysis steps and their settings is stored. These analysis versions can then also be used in the workflow to ensure consistent behavior.

Analysis versions can be created by creating an analysis model. Refer to Analysis Model

For the “SDS Extraction” analysis currently version 8 is active. This version stores all configuration of the analysis when the model was created on June 3rd 2025 at 11:01 PM.

Where it is Used

Analyses are used

stand-alone as part of the setup and training phase of a workflow and
as a workflow node to automate structuring documents or extracting data from documents

During the setup and training phase of a workflow the analysis is used stand-alone using the Analysis Document Viewer. In the analysis document viewer, the user can

add analysis steps
specify document annotations
preview the export assets

Once this is done, an analysis model can be created an the analysis can be used as a workflow node inside a workflow.

The analyses “SDS Mixture Extraction” and “SDS Substances Extraction” are used in the “SDS Workflow” to automate extracting data from safety data sheets

Getting Started

Create a new analysis either in the workflow editor or by clicking on the + next to Analysis in the left side panel . Give it a descriptive name.
Create or navigate to a Collection where you have your documents for this use-case
Right-click on a document and choose
With the right-click option “Open in ‘SDS Extraction’ the selected document can be opened in the analysis document viewer of this analysis
Use the analysis document viewer to setup your analysis and start annotating the Annotation Steps for a few of documents
Click on the analysis in the left side panel to navigate to the analysis model page and create a model based on the previous collection
Use theModel Evaluationpage to evaluate your model and activate it
Activate the model to use it in the analysis document viewer
Open an un-reviewed document in the analysis document viewer and use the “Run Analysis” button to validate the outcome
Once everything looks good, start automating by using the analysis in the workflow editor.
The analyses “SDS Mixture Extraction” and “SDS Substances Extraction” are used in the “SDS Workflow” to automate extracting data from safety data sheets

How To Use

Sub-Pages

For any questions you can contact our support team.