Data Language
ProductsMaturity ModelsBlogNewsAboutContact
Floating Cubes
Problem

Lack of a training corpus for AI systems

Many organizations intend to implement an AI automation service, but find that they do not have an adequate training corpus. This can be as a result of suboptimal information management, or low quality metadata among other issues.
Improved by these Interventions
Introduce tooling that creates structured metadata
Tools that enable metadata managers to create and manage structured metadata that conforms to a model. This enables more use cases for both human and machine users to leverage the metadata.
Generate a "Silver Corpus" for AI training
A "Silver Corpus" is an efficient bootstrapping approach to creating a precursor to a "golden corpus" in order to train an AI process efficiently.
Technologies used
Machine Learning (ML)
Training Corpus
Prevented by these Capabilities
An integrated information backbone
Data Language specializes in high-value information systems design and governance. A strong and highly contextual information backbone is essential for delivering and maintaining competitive digital products.
Metadata is part of a feedback loop to optimize AI RPA training
A business' core assets and associated metadata are part of a continuous optimization loop, where SMEs confirm or correct predictions, and the system learns from this.
Automated tagging integrated with knowledge graph
Media assets include structured, descriptive metadata
This capability is where a business' assets include metadata that is more than just descriptive text or stand-alone tags. The metadata ("tags" or "codes" or "annotations") includes structure that tells machines and humans what type of metadata it is, and what it means. It might also be connected via a graph to other metadata terms.
Caused by these Problems
Poor quality content tagging
Metadata holds no value for human or machine processes
Seen in these Maturity Models
Digital Media Metadata Maturity Model
Stage 1: Basic Metadata - Maturity Stage
Assets have simple, unstructured "Tags" used to power free-text search and can aggregate/group the assets. This level offers no information for machines or humans beyond free-text search and basic collections.
Digital Media Metadata Maturity Model
FAQs
Why does AI require a training corpus?

A training corpus is required to help an AI service understand "what good looks like" - i.e. what patterns to follow, and what the predictions or outputs should be formed.