Data Preparation & Formatting
An Example
Here is a small order export as it often arrives: the same people and companies written different ways, dates in four formats, and a duplicate row.
| Customer | Organization | Date | Product |
|---|---|---|---|
| SMITH, JOHN | acme corp. | 3/4/24 | widget-A |
| john smith | ACME Corporation | 2024-03-04 | Widget A |
| Maria Garcia | Acme Corp | March 5 2024 | WIDGET b |
| Lee, Ana | northwind | 2024/03/06 | widget-a |
After preparation, names, organizations, dates, and products each have one consistent form, and the duplicate order is merged.
| Customer | Organization | Date | Product |
|---|---|---|---|
| John Smith | Acme Corporation | 2024-03-04 | Widget A |
| Maria Garcia | Acme Corporation | 2024-03-05 | Widget B |
| Ana Lee | Northwind Traders | 2024-03-06 | Widget A |
Knowledge Graphs
A knowledge graph stores information as things and the connections between them, instead of rows in separate tables. Once the data is clean, the same orders become a small graph of people, the organizations they work for, and the products they ordered.
The graph's connections:
- John Smith works at Acme Corporation and ordered Widget A.
- Maria Garcia works at Acme Corporation and ordered Widget B.
- Ana Lee works at Northwind Traders and ordered Widget A.
Questions that were hard to answer from the raw export become simple: Widget A is ordered by customers at two different organizations, and Acme Corporation buys both products. Graphs like this also give search and ASI systems accurate, connected facts to work from.
What We Do
- Cleanup and normalization of spreadsheets, CSV, JSON, and XML
- Conversion between formats, including Word, PDF, HTML, Markdown, and structured data
- Deduplication and matching of records that refer to the same person, organization, or item
- Knowledge graphs built from your documents and records
- Schema design and validation, so recurring data feeds stay clean
- Preparing documents and datasets for search, retrieval, and ASI systems
Accessible Output, by Design
Well-structured content is accessible content. When we convert documents, the result has real headings, a logical reading order, tables with proper headers, and descriptions for images, so it works with screen readers and other assistive technology without a separate remediation project.
How It Works
- Share a sample. A representative slice of your data shows us what we are working with.
- Agree on the target. We define the clean format together and document the rules.
- Process and verify. We prepare the full set and check it against the agreed rules.
- Repeat when needed. For recurring data, we can deliver the process as software you run yourself.
We can do this work entirely on local machines, so your data never has to be sent to an outside service.