A century of telephony: digital capture of British telephone directories, 1880-1984

Figures & data

Table 1. Numbers of unique directories and of subscribers available from the digital scans from 1881 to 1951.

Year	Pages	Columns	Unique Directories	Estimated Total Number of Records
1881	535	535	7	12,847
1891	1,145	1,145	3	32,471
1901	1,027	1,027	5	46,064
1911	5,612	10,938	11	717,745
1921	5,317	10,634	7	645,811
1931	12,081	28,387	11	2,199,345
1941	7,743	22,581	8	1,719,772
1951	12,605	37,815	10	2,760,380
Total	46,065	113,062	62	8,134,435

Table 2. Examples of the different types of telephone directories that were made available by BT archives for this research.

BT roll name	Type of content
bt_900008_box01_1881_jan_001	Telephone directory containing records of all residential and business telephone subscribers from one city (London), published in January 1881.
bt_900011_box01_1881_jul_001	List of Professions & Trades - only lists subscribers which are offering their professional services or as tradesmen.
bt_900591_box74_1911_jul_001	Full listing split across multiple directories – each directory contains information for subscribers in Metropolitan London whose names begin with certain letters: A-H (900591) and H-Z (900592). This structure occurs with later London directories.
bt_900592_box74_1911_jul_002
bt_900789_box105_1941_oct_001	Regional directories – beyond major cities whose numerous subscribers are enough to constitute a directory of their own, most other records are listed by region which loosely correspond to historic county delineations.

Figure 1. Flowchart of the data pipeline for processing scanned telephone directory records.

Table 3. Classification of directory pages prior to digital encoding of names and addresses.

Page type	Information contained
Advertisements	Full-page advertisements of products or services, containing text that is often accompanied by images
Blank	–
Introductory	Introductory information for telephone directory end-users, for instance “How to use this directory” guides
Maps	Maps (typically hand-drawn) of telephone service provision, such as the location of telephone exchanges and call offices
Subscriber Records	Alphabetically or spatially organised lists of telephone subscriber names, telephone numbers and addresses
Telephone Exchange Information	List of addresses of the main telephone exchanges from which telephony services are provided
Trades Lists	Telephone subscription information organised by profession rather than alphabetically or by subscriber location

Figure 2. Use of binarization (right) to increase contrast of a scanned image (left). Example shows a scanned archive page from 1901.

Table 4. Some functions of the ‘teldiR’ package and examples of their use cases.

Category	Examples of use case
String operations	split_strings and trim_string have self-explanatory uses but also allow users to restrict the querying of text patterns to only the first X characters or last X characters of a string (where X is a number) collapse_RowsUpIf allows users to conditionally merge two or more rows of data, for when a record in the original directory occupied multiple lines
Process outputs of text capture	gen_fileLST lists all text files in a given system folder for easy importation and allows users to easily select subsets of the data, relevant for instance if they are interested in records (of a town/city) corresponding to a defined range of file numbers
Import and export tabular records	readin_adrDT imports already processed records (for viewing or further modification), with the option of choosing random samples from the dataset for troubleshooting or trial of new code export_fullDT exports processed records, naming the output file with a consistent format and automatically appends the time of export for purposes of versioning
General utility	repl_streetAbbrevs searches address fields for common abbreviations in throughfare names and replaces them with corresponding long versions serialise_recIDs generates a unique 11-digit identifier for each record which helps when they are to be matched to other telephone records from different years, or to census records

Table 5. Key differences between common measures of inter-string distance implemented in R.

String Matching Method	Allowable String Operations
	Deletion	Insertion	Substitution	Transposition
Hamming * requires strings of equal length	N/A	N/A	Yes	N/A
Levenshtein^†	Yes	Yes	Yes	No
Optimal String Alignment (OSA) ^†	Yes	Yes	Yes	Yes, once
Full Damerau-Levenshtein^†	Yes	Yes	Yes	Yes, multiple
Longest Common Substring (LCS)	*extracts the longest substring common to both strings, keeping the order of characters the same
^† for these methods, the weight that each kind of string operation carries in influencing inter-string distance can be adjusted manually

Table 6. Most common detected encoding errors (London directories).

Actual Character	Common Misdetections
S	‘5’, ‘8’
E	‘B, ‘F’, ‘H’, ‘13’, ‘18’
C	‘G’, ‘O’, ‘0’
.	‘’, ‘’, ‘-’

Figure 3. Composite map (left) of London’s civil administrative boundaries in 1881 (right, top) and 1914 telephone exchange locations (right, bottom).

Figure 4. Locations of 1901 telephone subscribers that were geocoded in (a) London, (b) Manchester and (c) Glasgow.