A paper from the School of Underlying Conditions and the School of Emergent Priorities shows that a classifier of field remarks does better when told how long each remark has been sitting there.
Three decades of notes beside groundwater readings are traced to the visit on which each was first typed
Most of what is written beside a groundwater reading was written on some other day. That is the starting point of Cap Seized, Use Stilsons: Carry-Forward Duplication and the Classification of Field Remarks on Manual Water-Level Readings, by Dr Vigdis Emenike, Senior Lecturer in the School of Underlying Conditions, and Dr Anouk Mensah, Postdoctoral Fellow in the School of Emergent Priorities.
The paper examines 143,906 remarks left by field technicians across a regional monitoring network, and reports that 68.6 per cent repeat, word for word, the remark entered at the previous visit to the same bore. One was carried through 263 visits in succession. Systems built to sort such remarks scored as high as 0.97 when tested in the customary way and 0.82 when asked only about remarks they had not already met.
I typed a number of these remarks myself, some while ago, and it was a curious thing to be introduced to them again as data. They have kept rather better than I have.
— Dr Vigdis Emenike, Senior Lecturer, School of Underlying Conditions
The authors followed every remark back to its first appearance, aligned each visit’s entry against the last, and anchored a flagging capability in what a technician changed on the day and how long the remainder had been left alone.
The University has a standing interest in things that persist, and is glad to see one of them given a depth and a date. “A sentence nobody has got round to deleting is still doing a job for somebody,” said Dr Mensah. “Finding out which job has been a pleasure, and I would not want to say it is finished.”
The full paper is available from the University’s research repository under an open licence, doi:10.5555/slop.ts5lrk.