Tracing fifty years of Setswana language change
A corpus turns scattered documents into evidence of how language behaves. For Setswana, also known as Tswana, a collection spanning roughly fifty years can show which words entered routine use, which meanings narrowed, and how spelling settled alongside print, broadcasting and digital communication. Studying Setswana language change over 50 years therefore produces a historical record rather than a single snapshot.
This evidence supports lexicography because dictionary writers need dated attestations, frequency information, context and a sense of variation between speakers, places and genres. It also reveals the uneven effects of language contact, translation and literacy. A sustained corpus lets researchers ask when change began, how quickly it spread, and whether present usage has a long local history or a recent borrowed form.
Why a fifty-year corpus matters
A useful design divides the collection into comparable periods: the 1970s, 1980s, 1990s, 2000s, and the 2010s to the present. Each slice should cover newspapers, books, letters, radio transcripts, school materials, religious writing and online texts. Counts can be normalised per million words, while keyword searches reveal contexts behind each token. Time changes the question from “What is used now?” to “How did usage arrive here?”
Balance matters more than sheer size. A corpus dominated by formal newspaper prose would exaggerate editorial vocabulary and understate conversation. Oral histories and everyday correspondence can counter that bias, although transcription may preserve features that formal editing removes. Digital lexicography offers useful models for maintaining such evidence, as discussed in The Future of Setswana Lexicography in a Digital World, but stable metadata and deliberate sampling remain essential.
Building a diachronic Setswana collection
Source selection begins with questions about coverage, not collection volume. Botswanan and South African archives can supply newspapers, church records, correspondence and school texts, while radio archives capture spoken material that print misses. Personal papers may preserve usages that editors standardised. Each item needs a date, place, genre and provenance so researchers can distinguish historical stages and avoid treating one voice as the language.
Papers must be scanned, corrected and checked against the originals, while born-digital documents need format-specific extraction. Optical character recognition often performs poorly on older type, faded ink and Tswana orthography. Corpus builders should retain the uncorrected layer for auditability and a normalised layer for searching, but normalisation must not erase spelling variation that may itself be evidence of change.
Australian collections in Sydney and Melbourne can hold migrant correspondence, community publications and institutional papers that illuminate usage outside southern Africa. Permissions and privacy controls matter, especially for living speakers, and a published corpus must comply with the Copyright Act 1968 (Cth). Cultural ownership and agreed access conditions should be recorded alongside technical metadata, especially where community knowledge is essential to interpreting a text.
Measuring vocabulary and semantic change
Frequency is the first measure, but raw totals can mislead. A token count records every occurrence, whereas a type count treats each spelling as one entry. Normalised frequencies permit comparisons across periods, and collocation analysis shows which words regularly keep company. A word that occurs once in a million tokens may still merit attention if its date and setting explain an emerging usage.
Semantic change is harder because the form can remain stable while its uses drift. Concordance lines allow a lexicographer to inspect a word across genres and decades, testing whether an apparent shift is real, temporary or tied to one publication. Etymological case studies, including The Story Behind the Word Morse in Setswana, demonstrate how individual histories can be reconstructed rather than guessed from modern intuition.
Statistical patterns still require linguistic judgement. A high count may reflect a prolific writer, a repeated news phrase or a policy debate, while a low count may conceal several specialised communities. In the Australian market for dictionaries and language technology, automated ranking can accelerate editorial work, but its categories and evidence still need review by people familiar with Setswana usage.
Spelling, tokenization and hidden patterns
Orthographic variation can make a stable word look like several entries. Capitalisation, hyphenation, loanword spelling and inconsistent spacing all affect tokenisation. Search systems should therefore retain original forms, map variants carefully and store normalisation rules in public documentation. Analysts can then distinguish genuine lexical innovation from mechanical differences between sources and typists.
Repetition also requires context. The account of rhetorical repetition shows how a writer can use recurrence for effect, so a frequency spike may reflect rhetorical strategy rather than durable change. Inspecting surrounding lines and comparing genres prevents dramatic-looking counts from being written into dictionary definitions as though every occurrence carried the same meaning.
Named people, places, abbreviations and compound forms create further traps. A tokenizer that splits or joins them incorrectly will distort both frequency and collocation. Corpus teams should test the pipeline with manually verified samples from different decades, then report the rules and known errors so future researchers can reproduce or improve the analysis.
Language contact and changing communities
Setswana has long interacted with English and neighbouring languages through trade, schooling, migration and media. This contact produces borrowing, calque, semantic specialisation and code-switching. A diachronic corpus can show which items remain productive, which suit formal domains, and which circulate only in bilingual conversation. It can also reveal whether apparent innovations are old regional usages entering wider circulation.
Experience from another setting offers useful comparative questions. The treatment of English loans in Japanese shows how borrowed vocabulary can settle, acquire local senses and create translation problems. Such comparisons should not flatten the history of Setswana, but they can sharpen attention to adaptation, register and the difference between a loan becoming dictionary-worthy and a word becoming part of daily speech.
In Melbourne, Sydney and Canberra, Setswana may appear in family messages, churches, classrooms, workplaces and social networks. Commuter routines and widely used messaging apps shape which varieties are recorded and who controls the archived record. Commonwealth agencies in Canberra use accredited translation and interpreting through the National Language Services scheme, showing how language provision has become part of everyday institutional life.
From corpus evidence to dictionary entries
Corpus evidence does not become a dictionary entry by frequency alone. Lexicographers must weigh date, region, register, speaker variation, semantic range and the quality of each attestation. A rare but well-supported word may matter more than a frequent spelling error. Citation selection should preserve enough context to let users judge whether an example represents ordinary use, a specialist context or a dated preference.
Historical layers also explain why particular expressions appear. The work on The History of the Setswana Bible Translation provides a valuable example because translated texts can shape vocabulary while retaining older forms, and because one influential genre can be echoed across later publications. Reading such material alongside newspapers and private writing helps separate broad change from the diffusion of a particular register.
An evidence-based dictionary can pair concise definitions with dated citations, frequency bands and links to corpus lines. Its editorial policy should state whether rare, obsolete, regional or nonstandard forms are included. Australian dictionary publishers and language-technology firms can apply the same transparent workflow to Setswana, while community reviewers help ensure that formal documentation does not erase local variation.
Preparing the next fifty years
A fifty-year baseline should become a monitored corpus rather than a finished database. New newspaper runs, broadcasts, books and web texts can be added at scheduled intervals, with methods kept stable enough for comparison. Periodic sampling also reveals changes in the archive itself, such as the growth of online sources and the decline of handwritten correspondence. Versioned releases would let users distinguish new evidence from revised analysis.
Computational models can identify candidate shifts, unusual collocations and spelling clusters, but interpretation remains central. Models trained on formal or English-centred data may misread code-switching or apply imported categories to Setswana. Cooperation among lexicographers, corpus specialists, speakers and archivists in Botswana, South Africa and Australia can connect computation with community knowledge, while careful access policies protect sensitive material.
What should remain clear is that Setswana change becomes intelligible when time, context and usage are considered together. A word count shows that something moved; a contextual record shows what it meant, who used it and why. Dictionaries benefit from both, preserving stable reference while making room for forms that are emerging, regional or socially marked.
The lasting lesson is that a corpus is more than a search engine over old words. It is a carefully governed historical instrument, capable of showing how a living language adapts to print, institutions, technology and contact. The memory of Setswana should therefore include both inherited vocabulary and the documented movement of language through the communities that speak it.