Utility to determine start/end years from textual expressions of year periods. Note this is a Python reworking of some earlier C# .NET project work (https://github.com/cbinding/timespans)

Archaeological dataset records often use a textual expression of dating rather than absolute numeric years for the dating of artefacts. These textual data values can be in a variety of formats and different languages. There can be prefixes present such as Circa, Early, Mid, Late - and suffixes such as A.D., B.C., C.E., B.C.E., B.P. that may influence the dates intended. This situation presents potential problems for temporal comparison of records in a single dataset, but also introduces wider data integration issues, as illustrated in the table of examples below:

Normalising such data can make subsequent search and comparison of records easier and more accurate. We can do this by supplementing the original data values with additional attributes defining the start and end dates of the timespan. This application attempts to match textual values representing timespans to a number of known patterns, and from there to derive the intended start/end dates of the timespan. For some cases the start/end dates are present and may be extracted directly from the textual string, however in most cases a degree of additional processing is required after the initial pattern match is made. The output facilitates better comparison of textual date spans as often expressed in datasets. Due to the wide variety of possible formats (including punctuation and spurious extra text or white space), the matching patterns developed cannot comprehensively cater for every possible free-text variation present, so any remaining records not processed by this initial automated method can be manually reviewed and assigned suitable start/end dates.

The output dates produced are derived relative to Common Era (CE). Centuries are considered to start at year 1 and end at year 100. Prefix modifiers for centuries take the following meaning (in this application):

*Note the boundaries of Early, Mid and Late overlap, suggesting that a level of approximation is intended when using such terms.

In the case of decades, centuries or stated tolerances, an offset is added or subtracted from the initial extracted year in order to interpret the overall extents of the year span being expressed. (e.g. 1540±9 = start year 1531, end year 1549)

For matches on known named periods (e.g. Georgian, Victorian etc.) the start/end years are derived from suitable authority list lookups.

Command:

python yearspanmatcher.py -i "{input}" [-l "{language}"]
python yearspanmatcher.py -i "Early 2nd Century" -l "en"

The textual timespan expression to be processed. The matching patterns employed are all case insensitive, e.g. 2nd Century AD and 2nd century ad would yield identical results.

The ISO639-1 two character language code corresponding to the language of the input data. This hints to the underlying matching process the most appropriate matching patterns to use. Languages currently supported are:

• Czech ('cs')
• Dutch ('nl')
• English ('en') [default]
• Italian ('it')
• German ('de')
• French ('fr')
• Norwegian ('no')
• Spanish ('es')
• Swedish ('sv')
• Welsh('cy')

If the language parameter is omitted or is not one of these recognised values then the default used will be en (English).

A suite of tests using the Python unittest framework are located under the tests directory. There are 315 unit tests in all, covering the various categories of year span textual expressions in each supported language.