My Gloucester Road project has come some way since I first wrote about this idea in April last year. It now has its own domain and the number of transcribed records from the Kelly's Street Directories is about 20% of the total.
Today I added some photographs of a section of the road known as Pigsty Hill. The properties have been empty for years so I wasn't in a hurry, but quite suddenly (to me) boarding has gone up and the buildings are being demolished to make way for housing. All the properties have a long history until their closure. At number 154, the upholster and his wife collected Bristol pottery, now in the Bristol Museum. At number 156, there has been a fish and chip shop for over 100 years.
The last occupant of number 156 were a very nice Chinese family who ran the Diamond Wok. Despite the fact that the business closed some years ago, Google finds about 40 pages on the internet, confidently enticing customers to this chinese takeaway. One of the worst is http://www.placesto-eat.com/diamond-wok-bishopston-bs7-8nt/ Note the inapporiate images (for a chinese take-away) and the not-so-helpful location in Beckenham, London.
These are business directory sites which seem mostly to harvest address lists. Entries seem rarely to be removed. Neither the source of the orginal data nor the date of acquisition is given so it is impossible for the viewer to rate the accuracy of the data for themselves. Given the date and an estimate of the churn in a locality, we could deduce the likelihood of an entry being still correct after N years.
The internet often seems to be an idiot savant that never forgets but never checks. That's a good question for internet science I think: in typical business online directories, what are typical rates for both businesses listed but no longer active (false positives) and businesses active but not listed( false negatives) ? One suspects that online directories have much higher rates for both than the published street directories from which our historical data is being extracted. In addition, both kinds contain simple errors, a problem for transcribers.
The local business directories, such as gloucester-rd.co.uk, Love Gloucester Road and Bishopston Matters manage better. False postives seem lower whilst false negatives seem higher, perhaps because businesses have to submit their own details.
I think of the Gloucester Road Story site is a prototype for a different kind of site, a site where tiime and place are core, the main dimesions of the data space.
Time must be explicit not just so that historical date can be included, but so that data about current occupants can become history as new occupants appear.
Place must be explict because it allows location browsing. Addresses are not just strings, they are located geographically. This means that a visitor can see what is next door, what is across the road (although you can't do that yet!). Using both dimensions, we can imagine a virtual street view at any time past. However we need to collect a lot more data to achieve this goal.
A third, fuzzier dimension is the type of business enacted. Since the street directories lack a formal naming convention some recoding will be needed. but the goal of being able to visualise the decline in boot-menders and the rise of coffee shops will be worth the effort.
Provinence is important because visitors should be able to check the validity of data themselves. This is straightforward for transcriptions from street directories since the photocopied pages are on the site. Personal observation and interview are more problematic and need attribution to individuals.
I now understand why addresses are treated as strings and not full entities. I'd naively assumed that numbering of addresses was monotonic - the house next door going up the street has a higher number. Not so - there is a whole terrace of houses which have the same number as houses further down the street. I'd assumed all premises had numbers. Not so: Churches and public buildings typically don't. Added to this are the complications which I had anticipated such as the change of address with time due to road renaming (Gloucester Road was formerly called Horfield Road), re-numbering or rebuilding. I dont imagine solving this problem, just a hard ongoing fight.
The sustainability of the site will depend on the ability of the team of local enthusists to keep it going, but I think this model puts us in with a chance. The challenge is to complete the historical sample and then lay down another, say, 10 years of history. Well, you have to take the long view.