To account for the estimated costs in the creation and managing of an NLP/ML classifier or model, there are three key elements: the human resources required (manpower), the infrastructure costs, and the ongoing maintenance costs to sustain the new capability.
Estimating Resource Costs
While the complexity of NLP/ML classifier models varies heavily depending on the use cases, this estimation is based on the creation of a semi-complex NLP classifier. An example of this is sentiment extraction or entity detection.
The average effort for the creation of a semi-complex NLP or ML classifier can vary in size, but often can be estimated at a duration of 8 ‘sprints.’ A Sprint is a measurement within engineering teams of dedicated time to specific stories and generally is aligned with 2 week cycles. This brings our estimation of duration to 16 weeks from planning to production release. The usual team composition and costs that are most common seen are laid out below…
READ THE FULL ARTICLE
