Indeed. What you’re suggesting (by extension) is that text is simply another form of data. And until such time as text objects are immutable and uniquely addressable, elevating the science of text-centric applications is not likely.
Optimistically, I believe Coda’s underlying architecture provides exactly this functionality; it is simply not exposed in a feature set [yet].
Crafty workarounds, while perhaps possible, and may provide some feeling of achievement, do not begin to put a dent in the deeper objectives of treating text as data and deriving all the benefits envisioned.
A key aspect of transforming text objects into first-class data citizens is the simple idea that (for example) a paragraph may possess arbitrary meta-values. And meta data about text is largely a fleeting idea and especially non-existent in previous-generation text processing tools. The “arbitrary” meta data distinction I make separates attributes like formatting indicators from descriptive attributes such as an array of inbound links to a paragraph from other documents.
Treating text as data exhumes a number of ideal benefits including, but not limited to:
- Automated document assembly from abstract text components
- Semi-automated introduction of atomized fragments into new documents
- Automated classification based on meaning and attributes
- Tagging and classification based on AI processes such as entity extraction
- Discrete search and addressability
- Discrete metrics and analytics
- Discrete link creation and discovery between documents and text objects
Imagine a graph showing the relationships between documents based on user-stipulated relationships in contrast to a graph that depicts the relationships based on automatically-extracted entities, or relationships based on meaning, or encapsulated ideas.
Smart solutions depend on data. Once text is invited to the data party, the science of data can be used to elevate the science of documents.