Analyzing Setswana parliamentary language using a concordancer

Parliamentary language is rarely as straightforward as a transcript of policy. Members frame proposals, challenge opponents, quote authorities, and appeal to shared values in dense spoken form. In Australia, where federal debates are recorded in Canberra and published through parliamentary Hansard, readers often encounter this language as a cleaned-up text. A concordancer changes that experience by showing how a word or construction behaves across actual utterances rather than presenting isolated dictionary examples.

For Setswana, this approach opens a route into debates shaped by oral performance, political convention, and contact with English. It allows a lexicographer to distinguish a recurring technical term from a ceremonial phrase, and a translator to judge whether a formal English equivalent preserves the force of the original. The analysis is valuable only when the corpus, tokenisation rules, and historical context remain visible behind every match.

From speeches to searchable evidence

A concordancer is best understood as a context engine. It retrieves occurrences of a target item and sorts them into keyword-in-context, or KWIC, displays, usually with the matched word centred among its neighbours. Applied to speeches, committee hearings, questions, and replies, these displays reveal how speakers nominate institutions, refer to citizens, concede a point, or signal disagreement. A simple frequency count might show that a noun is common; the surrounding lines can show whether it names an office, a policy instrument, a place, or a political opponent.

The wider history of public authority matters here. The article on religious and political rhetoric describes how formulaic invocations can legitimise speakers and connect political action to a moral order. Parliamentary debates in Botswana and South Africa can contain comparable rhetorical moves, although their institutions and audience conventions differ. Concordance lines let the researcher examine that resemblance linguistically instead of treating both traditions as interchangeable.

Building a defensible corpus

Corpus design comes before searching. The researcher must decide which parliamentary records form the period, whether proceedings include committee transcripts, and whether spoken interruptions are tagged. Metadata should preserve speaker, party, sitting date, chamber, procedure type, and document source. A question-answer exchange, a party-leader response, and a prepared ministerial statement should not be merged into one undifferentiated collection, because each has a different communicative setting.

In Australia, access to records is shaped by both routine and law. Parliamentary offices publish official transcripts, while the federal Freedom of Information Act 1982 can provide access to unlisted material; state and territory procedures differ. Copyright and database licences also affect what may be stored or redistributed. The professional profile is relevant context for this kind of work because it connects textual analysis with dictionary-making and African language preservation.

Text preparation needs comparable care. Optical character recognition errors, duplicated headers, page furniture, and inconsistent speaker labels can distort both counts and collocations. Original spelling should be retained in an archival copy, while a normalised working version can standardise whitespace and Unicode forms. Every editorial decision should be logged, including whether laughter, applause, or overlapping speech remains part of the searchable text.

Handling Setswana morphology and meaning

Setswana word forms do not always map neatly onto whitespace-delimited English units. Noun classes and agreement can enlarge a phrase, while verbal morphology carries tense, aspect, mood, and subject information. A concordancer that searches only one inflected form will undercount a lemma; one that matches too broadly may mix unrelated senses. Tokenisation, lemmatisation, and part-of-speech annotation therefore need to be tested against real parliamentary examples rather than assumed to transfer from English.

A practical query can preserve the surface token, search a lemma with its morphological tag, or restrict a multiword expression. Each produces different evidence. Searching polisi can expose collocates such as tsepamo or kagodimo when those forms actually occur. A phrase query for “the government” can similarly reveal whether speakers use puso, buso, or an English switch, and whether the choice changes the tone of accountability.

The display should still permit manual checking. A lemma can cross grammatical classes, and a tagger may confuse names, borrowed terms, and ordinary vocabulary. Corpus linguists often keep a core query, a tagged query, and a manually verified sample. That layered approach makes the analysis auditable: a reader can move from a general pattern to the exact parliamentary sentence and decide whether the interpretation is sound.

What concordance patterns reveal

Patterns of collocation show which concepts travel together. If a speaker repeatedly links puso with words concerning service, resources, or public trust, the cluster may indicate an institutional concept in active use. Colligation reveals a different layer: whether that noun tends to occur with possessive constructions, locative phrases, passive verbs, or expressions of certainty. Such grammatical habits often carry more argumentative weight than the bare topic word.

Frequency still requires restraint. Repeated wording may reflect party discipline, a standing agenda, or a transcription template rather than a national preference. Australian listeners will recognise a similar problem when ministers use the same polished formulation across question time in Canberra, or when newspaper reports compress a long exchange into one quotation. The Evidence Act 1995 governs evidence in Australian courts and related proceedings, while parliamentary committees have their own procedural rules; keeping those settings distinct prevents a legal finding from being inferred from lexical frequency. Concordance evidence describes discourse, not guilt, approval, or public opinion.

Language contact, history, and translation

Parliamentary Setswana contains vocabulary acquired through formal schooling, administration, media, and everyday contact with English and Afrikaans. The corpus may also preserve older words that remain current in some communities but have faded from formal speech. Searching a lemma together with date, speaker, and procedure can show whether a term belongs to inherited usage, a recent borrowing, or a ceremonial register. Translation must therefore consider time, institution, and audience as well as referential meaning.

The history of individual words adds another dimension. T.J. Otlogetswe’s article on the word kgomo shows how an entry can move beyond a short gloss into an account of cultural and semantic background. A parliamentary concordance can test that kind of history by locating kgomo in speeches about leadership, authority, or community. The surrounding English loans and Setswana expressions may reveal how older concepts are remade for modern political debate.

This approach has a place in contemporary Australian life even where Setswana is not widely spoken. Multilingual households in Sydney and Melbourne encounter English alongside many community languages, and labels on supermarket shelves at Woolworths or Coles illustrate how borrowing, regulation, and marketing interact. For many readers, parliamentary coverage arrives as a news bulletin during the morning commute or as social-media clips over a lunch break. In Canberra, official wording circulates even faster, making a transparent corpus useful to translators, teachers, and community language workers.

A reproducible research workflow

A defensible project begins with a written research question and a named corpus edition. For example, the question might ask how opposition speakers construct responsibility during budget debates. The researcher records the date range, document types, exclusions, and search fields, then deposits a sample of the data and concordance outputs with the analysis. Versioned software and settings make it possible to rerun the same query after corrections.

Quality control should combine automation with close reading. A second researcher can check a stratified sample of matches, disagreements can be reviewed, and ambiguous tokens can be assigned explicit reasons for inclusion or exclusion. Where proper names, quotations, or formulaic phrases create noise, filters should be published alongside the results. Access conditions and community permissions should also be recorded when transcripts contain sensitive testimony. The aim is to explain why particular matches were retained rather than erase difficult language.

For scholars seeking technical support or commissioned corpus development, the professional language services page provides a relevant point of contact. The practical takeaway is to preserve an auditable corpus, publish the query design, and require every concordance claim to remain traceable to a dated parliamentary source.