Page 404 - Contributed Paper Session (CPS) - Volume 4
P. 404
CPS2449 Louisa Nolan et al.
• Big Data is not a silver bullet, but we can create useful evidence, even
where outputs are not – or not yet – of the same quality as official
statistics
• building a new knowledge base for quality, not just of novel data
sources, but also of novel techniques, is challenging, but the principles
remain the same: understand user needs so that these are adequately
met, and explain quality clearly, so that users are confident in knowing
how and when to use new outputs; we may also need new approaches
to quality assurance, which give the subject matter experts confidence
in results from complex algorithms, such as the web application
developed for Optimus
• collaboration is key for data science teams, and that includes
collaboration with subject matter exerts and data experts, as well as
the ultimate users
• AI is also not a silver bullet, and human-in-the-loop approaches allow
humans to focus on oversight and quality assurance for difficult cases,
with AI processes relieving humans from the burden of repetitive
processing
• scale-up, from development and prototyping to production, requires a
different set of skills from development, and possibly a different team,
again, collaboration is vital in ensuring a smooth handover.
All this work is more fully described in the references (1,2,3).
References
1. Nolan L. et al. (2019), Faster indicators of UK economic activity,
https://datasciencecampus.ons.gov.uk/faster-indicators-of-uk-
economic-activity/
2. Pugh D., Williams S. & Johnson C. (2019), Understanding the
characteristics of high growth companies using non-traditional data
sources, https://datasciencecampus.ons.gov.uk/projects/understanding-
the-characteristics-of-high-growth-companies-using-non-traditional-
data-sources/
3. Hopkins S., Clews G. & Eidukas A. (2019), Optimus – A natural language
processing pipeline for turning free-text lists into hierarchical datasets,
https://datasciencecampus.ons.gov.uk/projects/optimus-a-natural-
language-processing-pipeline-for-turning-free-text-lists-into-
hierarchical-datasets/
4. Research Output: Economic activity, faster indicators, UK,
https://www.ons.gov.uk/economy/economicoutputandproductivity/outp
ut/articles/economicactivityfasterindicatorsuk/previousReleases
393 | I S I W S C 2 0 1 9