Language science and language technology in Africa: A festschrift for Justus C. Roux
Summary (3 min read)
1. INTRODUCTION
- Currently, the large-scale assessment of language proficiency, particularly at higher education levels, is dominated by reading and writing tests because listening and speaking skills are thought to be too difficult to evaluate.
- The same methods can be employed for oral proficiency assessment, although it must be borne in mind that speech carries (among other attributes) the accent and gender of the testee, both of which can increase the possibility of bias.
- They are logistically more problematic, particularly when the results must be available in a short space of time.
- Being proficient in academic English is extremely important because even if they did not continue their study of English after the first year, the students would still Automatic assessments of oral proficiency and indicators of linguistic abilities 311 use the language in their academic subjects.
- The authors report on the correlation between human and automatic ratings for a test population of first year students, as described in the preceding paragraphs.
2. DO WE NEED LINGUISTIC THEORY? RANDOMNESS AND SYSTEMATICITY
- Previous scholars entertained the view that word stress in BSAE is “assigned idiosyncratically, very often on the penultimate syllable, following the phonological rule in Bantu languages where this syllable is lengthened” (de Klerk & Gough 2002:361; see also Hundleby 1964:80-81).
- In his work, van Rooy (2002) argues that neither of the hypotheses can account for his data in a satisfactory way.
- This observation is central in the penult-hypothesis, and explanatory reference is made to the prosodic system of South African Bantu languages.
- Van Rooy (2002) observes that in BSAE the phonological shape of the final syllable also determines stress assignment, more specifically either a consonant cluster in the coda of the final syllable (thus syllable weight; see 1c) or the presence of an underlying diphthong in the final syllable (see 1d) leads to such syllable receiving stress.
- For the latter one needs to differentiate between two kinds of suffixes (or affixes more generally): on the one hand those which are incorporated into the stem for the purposes of stress assignment, called opaque suffixes by van Rooy (2002:153), and on the other hand those whose presence is ignored in stress assignment, called transparent suffixes.
3. REPRODUCABILTY: TESTING THE PREDICTIONS OF THE STRESS ALGORITHM
- In order to test if the predictions of van Rooy’s (2002) algorithm are borne out in new data, the stress patterns of polysyllabic content words in an already existing corpus of spoken Black South African English have been analysed.
- The corpus contained the speech of 14 speakers of Black South African English.
- Speakers were asked to read nine short paragraphs, containing between 35 and 65 words each.
- The nine paragraphs contained 79 polysyllabic content words which were subsequently transcribed with respect to their segmental and suprasegmental features by the author of the study and a student researcher.
4.1. Supporting the Stress Algorithm
- The first observation when comparing the actual realisations to the General South African English realisation is that the divergences between the two varieties in the segmental domain are considerable (mostly with respect to vowel quality and the manner of articulation of the rsound).
- Furthermore, van Rooy’s (2002) algorithm is confirmed in a variety of ways that will be presented and discussed in detail below.
- The fact that most words are realised with stress on the penultimate syllable is accounted for by the important generalisation in van Rooy’s (2002) algorithm, namely that stress is usually on the penultimate syllable.
- In both cases, stress would be predicted on the final syllable when based on the underlying form given that they are superheavy (VCC).
- (6) orthography underlying predicted realised promised prɒmɪsd proˈmist 'pɹɔmɪz (11), 'pɹɔmɪzd (2) boyfriend bɔɪfrend bɔɪˈfrent 'bɔɪfrɛnd, 'bɔɪfrɛn (10), 'bɔɪfɹɛn (3).
4.2. CHALLENGING THE STRESS ALGORITHM
- As the examples in (2) – (6) show, the new data that have been collected support the BSAE stress algorithm by van Rooy (2002) in a variety of ways.
- The reader might have noticed these instances in (3) and (6).
- (7) orthography underlying predicted realised a. divorced dɪvɔːsd diˈvosd diˈvɔs received rɪsiːvd riˈsivd riˈsit, riˈziv b. promised prɒmɪsd proˈmist 'pɹɔmɪzd (2).
- The examples in (11) show a principled restriction of van Rooy’s stress algorithm4.
- For all examples in (11), stress is predicted to occur either on the final or on the penultimate syllable whereas it is actually realised on the antepenultimate syllable in agreement with the Standard English stress pattern.
5. DISCUSSION
- The previous section has presented new data from a corpus of read speech which was evaluated against the BSAE stress algorithm proposed by van Rooy (2002).
- It could be shown that the algorithm accounted for many of the occurring stress realisations and that the corpus thus provided support for all of the assumptions made by van Rooy (2002).
- It also became clear that the algorithm – not surprisingly – cannot account for all the data.
- The logical step forward would now be to propose a modification of the algorithm that would do exactly that.
- Instead the discussion section wants to raise two concerns that prevent me from doing so: these concerns relate to the representation of word prosody in language contact, and the role of grammar and frequency in language.
5.1. Word-Prosodic Systems in Contact Languages
- By definition, mesolect speakers use a variety which lies between the two languages involved in the language contact situation (Bickerton 1971).
- Developing a stress algorithm for mesolect speakers of BSAE in fact assumes a stress system, an assumption which contradicts the above-mentioned assumption.
- Even in simultaneous French-Spanish bilinguals the phonological representation of stress differs from that of monolingual speakers of the languages.
- Gussenhoven and Udofot (2010; see also Gut 2005) investigate sentence prosody in Nigerian English, a variety arising out of the contact between a stress language and a West African tone language.
- Also the literature on the acoustic realisation of sentence stress suggests that contact languages can creatively develop prosodic systems which differ from the prosodic systems of both the substrate and the superstrate languages.
5.2. The Role of Grammar and Frequency in Language
- In order to disentangle the two general mechanisms, the welldocumented influence of usage statistics needs to be controlled for in linguistic studies that address this question so that results can be interpreted as giving evidence about grammar per se.
- A frequent experimental paradigm that controls for usage statistics is the use of nonsense words, i.e. madeup words that do not exist in the language under investigation.
- Recently, Zonneveld (2010) reanalysed Van der Pas et al.’s data as showing evidence that speakers of Tswana English show an interim grammar in which final VC-syllables are extrametrical and thereby open up the possibility of antepenultimate stress.
6. CONCLUSION
- Using the general theory of lexicography to plan a dictionary gives lexicographers the opportunity to opt for existing lexicographic procedures or new procedures, but also to combine existing and new procedures by using existing procedures as a basis for innovative adaptations and applications.
- The way in which numerous multiword terms have been entered into the macrostructure by means of a variety of procedures resulted in different types of sublemmata.
- In this regard the WGW has applied existing theoretical guidelines in an innovative way to ensure the best possible and wide-ranging macrostructural coverage.
- Here a contemplative approach to lexicography offers the basis for innovative transformative approaches.
- LANGUAGE SCIENCE AND LANGUAGE TECHNOLOGY IN AFRICA: A FESTSCHRIFT FOR JC ROUX 268.
Did you find this useful? Give us your feedback
Citations
1 citations
References
19 citations
"Language science and language techn..." refers background or methods or result in this paper
...The findings confirmed that the algorithm can account for many of the observed stress patterns and thereby refutes, in line with van Rooy (2002), the previously held impression that stress assignment in this variety is either idiosyncratic or restricted to the penultimate syllable....
[...]
...(4) orthography underlying predicted realised lions laɪəns 'laɪens 'laɪons (7), 'laɪəns (7) Van Rooy (2002) claims that in order for a syllable to be superheavy and thus to attract stress it needs to have either a diphthong in the underlying form followed by at least a single consonant or a vowel…...
[...]
...It could be shown that the algorithm accounted for many of the occurring stress realisations and that the corpus thus provided support for all of the assumptions made by van Rooy (2002)....
[...]
...Starting out from the basic requirement that all lexical content words need to have one primary stress, van Rooy (2002) isolates three factors that account for almost all of the data....
[...]
...The previous section has presented new data from a corpus of read speech which was evaluated against the BSAE stress algorithm proposed by van Rooy (2002)....
[...]
18 citations
2 citations
Related Papers (1)
Frequently Asked Questions (15)
Q2. What future works have the authors mentioned in the paper "Language science and language technology in africa" ?
These issues remain for future work. Bearing in mind the numbers of less resourced languages which do not yet have advanced technological infrastructure, there will continue to be a place for such tutorial educational methods for a long time to come.
Q3. What is the common experimental paradigm that controls for usage statistics?
A frequent experimental paradigm that controls for usage statistics is the use of nonsense words, i.e. madeup words that do not exist in the language under investigation.
Q4. What is the meaning of opaque suffixes in BSAE?
Other examples of opaque suffixes are the nominal suffixes -ment, -ion, -or and -ature as well as the adjectival suffixes -ant and -able.
Q5. What are the main parameters of the current corpus study?
Corpus studies such as the current one and its precursor (van Rooy 2002) are useful to delineate relevant parameters such as morphological complexity, underlying vowel quality and surface consonant clusters.
Q6. Why is the penultimate position of a word preferred?
The preference for the penultimate position has been claimed to be due to the fact that in the South African Bantu languages, the penultimate syllable of a word is predictably lengthened.
Q7. What is the underlying syllable in van Rooy’s algorithm?
Note that only underlying diphthongs but not long vowels as in (3) constitute a long syllable nucleus in van Rooy’s BSAE stress algorithm.
Q8. What is the phonological representation of a specific language’s word prosodic system?
Work by Dupoux and colleagues on the perception of word stress in foreign language acquisition and bilingualism has shown that the phonological representation of a specific language’s word prosodic system is very stable in language contact.
Q9. What is the reason for the manipulation of different acoustic parameters in Nigerian English?
A manipulation of different acoustic parameters might, at least partly, be due to the fact that these African Englishes are varieties of English that arose in contact with African tone languages, and tone languages use F0 for lexical and grammatical distinctions.
Q10. How many participants in Van der Pas et al.’s study produced stress patterns comparable?
Three to ten out of the 50 participants in Van der Pas et al.’s study produced stress patterns comparable to a Canadian English control group, tested in Pater (1997).
Q11. What is the effect of a reanalysis of Van der Pas et al.?
Zonneveld (2010) reanalysed Van der Pas et al.’s data as showing evidence that speakers of Tswana English show an interim grammar in which final VC-syllables are extrametrical and thereby open up the possibility of antepenultimate stress.
Q12. How does this paper attempt to contribute to the expansion of a theoretical model?
Using a contemplative approach, this paper endeavours to contribute to the expansion of a theoretical model by making the procedures that are discussed accessible to future lexicographers.
Q13. How many times did he serve as the Chair of the Department of African Languages?
He was the Chair of the Department of African Languages at Stellenbosch six times (1982, 1986, 1988, 1992-1993, 1997-1999, and 2000).
Q14. What was his role in the development of the NCHLT-RMA?
Based on his experience in the field of resource management and his collaboration with international language resource projects, he was also appointed co-ordinator of a Task Team to develop a Blueprint for a Resource Management Agency for the National Centre for Human Language Technologies (NCHLT-RMA) in 2010.
Q15. What does Van der Pas et al. (2000) show about stress in Tswan?
their results show that speakers of Tswana English often do stress the antepenultimate syllable: the majority of participants produced antepenultimate stress on 5 out of 16 words, such as tadimet, kadowtet, and kapistratson.