The Silchester article makes a number of references to wells, including the following discussion from Period 2: Early Roman Timber Building 2.

'Room 5 was floored with silty clays and gravels (5479, 5401, 5777, 5407, 3704, 3972), some of which had subsided into an earlier well 5791' and also from the hearth extract above ... 'This hearth had been cut through by Late Roman well 1044.'
Close up of interface

Some new issues surface when exploring Demonstrator scenarios around the concept of wells, in part due to the broad and ambiguous nature of the term itself. When building a query for Contexts of this type, various potential concepts are suggested by the Demonstrator, which raises the archaeological issue of whether the search should distinguish the well-cut from information relating to when the well was constructed (e.g. construction pit) and generally in use, or whether to include also the well fills and backfills and methods of construction. Results reveal some differences of practice (and terminology) by projects, for example in assigning data to a well as the overall feature, as opposed to a sub-division of data between the cut of the well and its fills. We consider some of the different cases:

Context type = well (for RRAD & LEAP equates to well cuts)
Returned = 100+ results: 46 RRAD, 100+ MoLA ROP, 1 LEAP, 100+ OASIS 

The LEAP result is context 1044 (5791 is returned by well: construction pit). Again, more advanced semantic search features could overcome some of these differences in practice by providing an option to expand the search automatically. In addition, section 4.1 considers whether physical relationships should be included in the model.

Where a term (such as well) has a variety of meanings and uses, we can encounter 'false positives' in the NLP results, a common issue with text-based search techniques. In this case, with the current version of the information extraction system, OASIS results include at least eight false positives, due to matches on well built or well-preserved, etc. This is further discussed in section 4.5.

Context type = well:fill 
100+ results: 1 LEAP, 1 MoLA ROP, 100+ RRAD

The above search exposes differences between results returned for the (previous search) construction phases of the feature and those returned for deposits that backfilled the well.

Group type = well
Returned = 100+ results: 83 RRAD, 18 MoLA ROP

and also note

Group > contains Context > Context Type = well
Returned = 68 results: 32 RRAD, 1 MoLA HHN99, 35 MoLA ROP

Depending on the recording practice, a search at group level might overcome these issues to some extent, since discussion at group level tends to give an overview of what was found in both the construction and backfill deposits from the well.


© Internet Archaeology/Author(s)
University of York legal statements | Terms and Conditions | File last updated: Mon July 18 2011