- A single set that contains both semi-structured and unstructured documents (different document types).
- A single document that contains mixed structure — for example, an unstructured contract with embedded tables, titles, headers, or footers.
Pick a scenario
Each scenario below shares a common pattern; only the activities you add to the processing flow differ.
Common workflow
1
Create a Document skill
Open Advanced Designer and click Create Document Skill on the start page.
2
Upload documents
On the Documents tab, upload the documents you’ll use to set up the skill.
3
Define fields and label
On the Fields tab, create and configure the fields the skill will extract. Label documents in the Reference section.
4
Add and configure activities
On the Activities tab, add the activities for your scenario (described below). Open each activity in the Activity Editor to configure and train it.
5
Test and publish
Click Test Skill Using Selected Documents to evaluate results. When the results are good enough, publish the skill.
Semi-structured and unstructured documents in one set
Use this scenario when one Document skill must process both semi-structured and unstructured documents — both belong to the same logical type and share the same set of output fields. Classify each document with a Classify By Text and Image activity, which combines text and geometry to handle low-quality images and documents that differ only by graphic features (signatures, seals). For best results, upload a roughly equal number of documents for each variant so the classifier has balanced training data. Then branch the flow with an IF activity:- Process semi-structured documents with a Fast Learning activity.
- Process unstructured documents with a Segmentation activity followed by a Deep Learning for NLP activity.
Table cells with fields embedded in cell text
Use this scenario when you need to extract specific values from inside table cells in semi-structured documents — for example, a borrower’s name and partial address embedded in a Closing Disclosure cell. Extract the cell as one block of text with a Fast Learning activity, and then run NLP activities on that block to pull out the embedded fields:- Named Entities (NER) activity for entities like names and organizations.
- Address Parsing activity to break addresses into components.
Unstructured documents with tables, titles, headers, or footers
Use this scenario for documents that are mostly unstructured (for example, contracts) but contain embedded semi-structured fragments such as tables, titles, headers, or footers. Detect plain-text paragraphs with a Segmentation activity and detect the semi-structured fragments with an Extraction Rules activity. Once each fragment is isolated, use the appropriate activity to extract its fields.