Written by Helen Caple

In a recently published study, I examined the processes (verbs) associated with group-based identity labels (like we, they, Australians, citizens) for self-representation in historical newspaper texts. The study corpus was small and exhaustive of one Australian newspaper, which allowed for detailed, qualitative analysis of transitivity. Transitivity ‘is concerned with a coding of the goings on: who does what in relation to whom/what, where, when, how and why’ (Hasan, 1988: 63), and the focus is on the clause as representation. Decoding what is going on in a text requires analysis of the processes (verbal groups or ‘goings on’), participants (nominal groups or who or what is involved in the process) and their position in relation to the process, and the circumstances (adverbial groups and prepositional phrases or where, when, how and why) in which this is all taking place.

Since the study was a corpus-assisted discourse analysis, I was also interested in examining patterns across the corpus in terms of most frequent usage and meaning. To achieve this, I categorised the results from the manual transitivity analysis into semantic categories using the ATAP Semantic Tagger (Jufri and Sun, 2022), which makes use of the Ucrel Semantic Analysis System (USAS) to categorise words in texts (see Archer et al. 2002).

The ATAP Semantic Tagger comes in the form of a Jupyter notebook (developed by the Australian Text Analytics Platform) and is available for anyone to use via GitHub. The tool automatically categorises and annotates words or multi-word expressions based on their meaning class, as further explained in this blog post. For example, the general meaning class of Q Language and Communication is sub-divided into categories such as Q1: Communication and Q2: Speech Acts, and entries are further sub-classified within these categories. In my dataset, the most frequent verbal processes and their semantic categories were Q2.2 Speech acts, as shown in Table 1.

Semantic categories	Q2.2 Speech acts	Q 2.1 Speech: Communicative
Number of instances	86	13
Most frequent processes	ask, call (upon), appeal, refuse, advise, advocate, recommend	say, inform, speak, state, discuss

Combining transitivity analysis with semantic categorisation offered several advantages for my study:

• It allowed me to triangulate my findings and verify the manual transitivity analysis.
• It alerted me to the potential misstep of making assumptions about the roles of participants in the clause.
• The patterns revealed through the semantic categorising confirmed my hypothesis that the investigated newspaper operated very much from an activism frame rather than focusing on news reporting. For example, as seen in Table 1, the verbal processes fell overwhelmingly into the semantic category of Q2.2 Speech acts (e.g. asking, calling for/upon, advocating and recommending), which one might associate with opinion writing, rather than the usual reported speech (e.g. say, state categorised as Q 2.1 Speech: Communicative) one might expect to see in a typical hard news story.

While this study focused on a very small, specialised corpus of newspaper writing, there is scope to apply the ATAP Semantic Tagger to much larger corpora, combining it with transitivity or other types of qualitative analysis.

References

Archer, D., A. Wilson and P. Rayson. 2002. Introduction to the USAS Category System. (October 2002.) Benedict Project Report.

Hasan, R. 1988. ‘The analysis of one poem: theoretical issues in practice’. In D. Birch and M. O’ Toole (eds.), Functions of Style, pp. 45–73. London: Pinter.

Jufri, S. and C. Sun. 2022. Semantic Tagger (v1.0). Australian Text Analytics Platform. (Computer software.) Available online at: https://github.com/Australian-Text-Analytics-Platform/semantic-tagger.