A Corpus-Based Analysis of Mixed Code in Hong Kong Speech

2012 International Conference on Asian Language bear on A Corpus-based Analysis of Mixed Code in Hong Kong Speech John Lee Halliday Centre for intellectual Applications of Language Studies Department of Chinese, Translation and Linguistics City University of Hong Kong emailprotected edu. hk AbstractWe present a head teacher-based analysis of the utilize of complicated code in Hong Kong language. From transcriptions of Cantonese television programs, we identify slope dustup embedded within Cantonese utterances, and study the motivations for such code-switching.Among the m either motivations observed in previous research, we found that four alone banknote for much than 95% of the pulmonary tuberculosis of face talking toing to in our row data across genres, genders, and age groups. We performed analyses everyplace more than 60 hours of transcribed speech, resulting in one of the vauntinglyst empirical studies to-date on this linguistic phenomenon. Key wrangling-code-m ixing slope head teacher linguistics. code-switching Cantonese II. PREVIOUS RESEARCH I. INTRODUCTIONWhile Cantonese is the mother tongue for the vast legal age of the people in Hong Kong, English is to a fault spoken by 43% of the population 1, reflecting the citys heritage as a British colony. A healthful-known feature of the speech in Hong Kong is code-switching, i. e. , the juxtaposition of passages of speech belonging to 2 different grammatical establishments or sub-systems, within the same exchange 2. Specifically, in the case of Hong Kong, the two grammatical systems argon Cantonese and English.The former serves as the matrix language, and the latter as the embedded language, resulting in Cantonese sentences with English segments such as (example taken from 3) butt jointteen heoi3 canteen jam2 caa4 lets go to the canteen for eat Here, the English segment delivers only one word (canteen), but in general, it can be a whole clause. We will use the general margin code-s witching rather than the more specific term code-mixing, which refers to switching below the clause level, even though most English segments in our star thusly contain only one or two wrangling (see sidestep 3).T present is already a large body of literature devoted to the study of Cantonese-English code-switching from the theoretical linguistic point of view 3,4,5. This paper investigates the motivations fag the use of complex code, on the basis of a large dataset of speech transcribed from television programs. In air division II, we outline previous research on the motivations of code-switching, and discuss how our investigation complements theirs. In Section III, we limn our methodology for star construction, in particular the design of the taxonomy of code-switching motivations.In Section IV, we present an analysis of these motivations jibe to genre, gender and age. The first major framework for classifying codeswitching motivations in Hong Kong consists of two categor ies carpetbag and orientational 6. Central to this framework is the distinction surrounded by words in high Cantonese and low Cantonese. In everyday conversations, a vocaliser or sotimes cannot identify any word from low Cantonese to describe an object, institution or idea (e. g. , application form). employ a word from high Cantonese (e. g. , biu2 gaak3), however, would sound too formal and therefore stylistically inappropriate.In expedient mixing, the vocalizer resorts to an English word the mixing is pragmatically motivated. In contrast, orientational mixing is socially motivated. The speaker chooses to use English (e. g. , barbecue) despite the availability of uniform words from both low Cantonese (e. g. , siu2 je5 sik6) and high Cantonese (e. g. , siu1 haau1), since he perceives the put forward matter to be inherently more western. This dichotomy has been criticized as overly simplistic, because of the ambiguity in defining lexical and stylistic equivalents among low C antonese, high Cantonese, and English.Instead, a four-way taxonomy is proposed euphemism, specificity, bilingual punning, and the linguistic rule of economy 7. This taxonomy is indeed further extended, in a study of code-switching in text media 8, to include quotations, doubling, identity marking, and interjection. These categories will be explained in detail in Section III. While these compartmentalisation systems atomic number 18 comprehensive and well grounded, they do not per se convey any sense of the relative importance or distribution of the different motivations.Our goal is, first, to empirically verify the coverage of these classification systems on a large dataset of transcribed speech and, second, to give quantitative answers to questions such as Which kinds of motivations ar the most prominent? Does the range of motivations differ according to the speech genre, or to the speakers gender or age? We now turn our maintenance to the methodology for constructing and an notating a speech corpus for these research purposes. III. DATA A. Source Material Our corpus is constructed from television programs broadcast in Hong Kong within the last four years by Television Broadcasts expressage (TVB).The programs belong to a variety of genres, including two drama series, three current-affairs take the stands, a news program, and a talk of the town show. The news program, TVB News at Six-Thirty, carries the most formal register, containing mostly pre-planned 165 978-0-7695-4886-9/12 $26. 00 2012 IEEE DOI 10. 1109/IALP. 2012. 10 speech by the anchor. The current-affairs shows, Tuesday hide, Sunday Report and Hong Kong Connection, argon serious in tone but contain spontaneous discussions. The talk show, My Sweets, is about food and drink.It alike contains spontaneous discussions, but the topics tend to be lighter. Although pre-planned, the speech in both drama series, Moonlight plangency and Yes Sir, Sorry Sir, is arguably the least formal in register, d esigned to reflect natural speech in everyday life. Details of these TV programs are presented in Table 1. Table 1 Television programs that serve as the source material of our corpus. Genre class Length authoritative Tuesday Report ( ), 135 episodes affairs ), X 20 minutes Sunday Report ( Hong Kong Connection ( ) Talk 24 episodes My Sweets ( ) show X 30 minutesEuphemism When a Cantonese word explicitly mentions something that the speaker finds embarrassing, s/he might favor for an English word that contains no such mention. For example, to avoid the effeminate body part hung1 breast in the word hung1 wai4 bra, the speaker might prefer to use the English bra (all examples are taken from 7) bra tau3 bra gaak3 gaak3 A princess whose bra is visible Specificity sometimes an English expression is preferred because its meaning is more general or specific compared with its near-synonymous counterparts, 7 in every low or high Cantonese.For example, the verb to book means to make a reserv ation for which no property or deposit is required, which is more specific than its closest equivalent in Cantonese, deng6 to make a reservation. It is often used in sentences such as book ngo5 soeng2 book saam1 dim2 I take to book 3 oclock Principle of Economy An English expression may also be preferred because it is shorter and thus requires less linguistic effort compared with its Chinese/Cantonese equivalent. 7 While the word check in has two syllables, its Cantonese equivalent baan6 lei5 dang1 gei1 sau2 zuk6 check-in on a plane has six.The principle of economy is thus likely the reason rump compound code such as check-in nei5 check-in zo2 mei6 aa3 Have you checked in already? The taxonomy in 8 builds on the one in 7, further enriching it with categories2 below Quotation When citing text or somebody elses speech, one often prefers to use the original code to avoid having to perform translation. An example is lineal speech What do you think? jau5 go3 pang4 jau5 man6 ngo5 what do you think A friend asked me, What do you think? manifold Originally named Emphasis or avoidance of repetition 8, it will be referred to as two-bagger 9 here to make it explicit, as this category refers to English words that are embedded alongside Cantonese words that puzzle the same or nearly the same meaning. The purpose is to emphasize the idea or to avoid repetitions. In the following sentence, it serves as an emphasis 2 News Drama TVB News at Six-Thirty ( ) Moonlight Resonance ( ), Yes Sir, Sorry Sir ( Sir Sir) 5 episodes X 20 minutes 4 episodes X 45 minutes B.Data touch From the television programs listed in Table 1, all code-mixed utterances were transcribed, preserving the original languages, either Cantonese or English. Following standard practice, lend words are not see to ited to be mixed code in our context, all English words (e. g. , taxi) that have been adapted into Cantonese phonology (e. g. , dik1 si2) were excluded. The TV legends corresponding to eac h of these utterances are also put down as part of the corpus. These captions are in standard Chinese, rather than Cantonese.Furthermore, alignments between the Chinese word(s) in the caption and the English word(s) in the utterance are annotated. This information will be used in the classification of motivations. Finally, two kinds of metadata about the speaker are recorded gender (male or female) and age group (teenager or adult). C. Taxonomy of Code-Switching Motivations Our goal is to quantitatively characterize the motivations behind code-switching to this end, each English segment in the Cantonese sentences in our corpus is to be labeled with a motivation. Due to time constraint, this classification was performed only on the currentaffairs and talk shows.The expedient vs. orientational classification system is too coarse for our purpose. Instead, we adopted the taxonomy in 7,8 as our starting line point, indeed introduced some new categories to accommodate our data. The cat egories in 7 are1 1 A fourth category, bilingual punning, is excluded from our taxonomy. As may be expected, punning is rarer in speech, and is indeed not found in our corpus. Among these categories is identity marking, for mixed code that marks social characteristics such as social status, education status, occupation, as well as regional affiliation. 8 We found it difficult to objectively identify this motivation, and excluded it from our taxonomy. 166 precise good very good m4 co3 aa1 Very good, very good interposition English interjections may be inserted into the Cantonese sentence. For example Anyway anyway nei5 hou2 sai1 lei6 ak1 Anyway, you are awesome A significant count of mixed code in our corpus, however, still does not competent into any of the above categories. Most fall under one of two reasons, Personal Name and enter.We therefore added them to our taxonomy Register This is roughly equivalent to the expedient category in 6, but will be referred to as Register in this paper to make the motivation explicit. Sometimes, the speaker cannot find any equivalent low Cantonese word, but feels awkward to use a more formal high Cantonese word (e. g. , paai1 deoi3 caller). As a result, s/he resorts to an English equivalent instead. For example, party hoi1 ci2 laa1 ngo5 dei6 go3 party Our party is starting Personal Name It is common practice among Hong Kong people to adopt an English name.Although this phenomenon may be considered orientational codemixing in terms of the western perception 6, it is given its own category, because it is very specific and accounts for a substantial amount of our data. A typical example is Teresa, Teresa ngo5 dei6 zing2 dak1 leng3 m4 leng3 Teresa, did we make it nicely? D. Annotation Procedure We thus have a total of eight categories in our taxonomy of code-switching motivations. Five of these categories namely, euphemism, quotation, doubling, interjection, and personal name can usually be unambiguously discerned.Th e annotator, however, has often found it difficult to distinguish between specificity, register, and principle of economy. To maintain consistency, we adopted the following procedure. When an English segment does not fit into any of the fiver easy categories, the annotator is to decide whether it has the same meaning as the Chinese word in the caption to which it is aligned. If it is deemed not to have the same meaning, then it is assigned specificity. If it is equivalent in meaning, and the annotator cannot think of any equivalent in low Cantonese, then it is labeled register.Lastly, if there is a low Cantonese equivalent, but its number of syllables is bigger than that of the English segment, then the motivation is principle of economy. IV. ANALYSIS English segments in Cantonese speech (section A), then discuss the distribution of the categories of motivations, both overall and with respect to genres, genders, and age groups (section B). A. Density and Length of English Segments It is well known that English words are sprinkled rather liberally in the Cantonese speech in Hong Kong. We standard how the frequency of English segments varies across different genres.As shown in Table 2, the frequency correlates with the register of the genre (see Section III. A). In the drama series, the most colloquial genre, one and a half English words are uttered per minute on average. The talk show occupies second place, and the current affairs shows have slightly less frequent English words. In the news program, where the speech is preplanned, the anchor did not utter any English word. Table 2 The total number of Cantonese sentences containing English segments, and the total number of English words transcribed. The last column shows how often an English word is uttered.Program genre Drama Talk show Current affairs News sent with English 219 487 1495 0 English words 259 625 1995 0 Frequency (words/min) 1. 4 0. 87 0. 74 0 Second, we measure the length of the English segme nts. Table 3 shows that the vast majority of English segments contain no more than two words. Across all genres, more than 80% of the English segments consist of only one English word. This figure is comparable to the 81. 4% for text data account in 8. Table 3 Proportion of English segments with only one (e. g. , canteen) or two words (e. g. , thank you).Program genre Drama Current affairs Talk show One-word 85% 85% 81% Two-word 11% 11% 17% This section presents some preliminary analyses on this corpus. We first consider the frequency and length of B. Motivations for the use of mixed code A plethora of motivations have been posited for the use of mixed code in Hong Kong (see Section II). Applying our proposed classification system (see Section III. C) on our corpus of transcribed speech, we call for now to discern the relative prevalence of the various kinds of codeswitching motivations. Table 4 shows the distribution of these motivations in the current-affairs and the talk shows. Four dominant motivations chiefly register, but also personal name, principle of economy, and specificity are attributed to more than 95% of the English segments. This trend is the same across genres (current-affairs and talk shows), genders (see Table 6), and age groups (see Table 5). All other categories, including quotations, euphemism, doubling, and interjection, are relatively infrequent. Genres. Among the four dominant motivations, register the use of appropriately unaffixed words is the most frequent motivation in both the current-affairs and 167 talk shows.Its proportion, however, is significantly more marked (47. 4%) in the talk show than in current affairs (36. 4%), reflecting the more versed nature of the former. Table 4 statistical distribution of code-switching motivations, contrasted between genres. Motivation Current affairs Talk show Register 36. 4% 47. 4% Personal Name 26. 8% 24. 5% Principle of economy 19. 0% 17. 6% Specificity 13. 2% 8. 2% Quotation 2. 1% 1 . 0% image 1. 4% 0. 4% Interjection 0. 9% 1. 0% Euphemism 0. 3% 0% Age groups. Table 5 contrasts the distributions of code-switching motivations between adults and teenagers in the current-affairs shows 3 .As mentioned above, the four major motivations remain constant. However, teenagers are much more likely than adults to use English words to achieve more informal register (52. 4% vs. 35. 1%). They also tend more to opt for English to save effort (23. 8% vs. 18. 6%). Somewhat surprisingly at first glance, teenagers speak to others in English name less often than adults (2. 4% vs. 28. 8%) it turns out that in the conversations in our corpus, teenagers often prefer to address adults with the more formal Chinese names, likely out of respect.Table 5 scattering of code-switching motivations, contrasted between age groups. Motivation Adults Teenagers Register 35. 1% 52. 4% Personal Name 28. 8% 2. 4% Principle of economy 18. 6% 23. 8% Specificity 13. 1% 14. 3% Quotation 1. 9% 4. 0% Do ubling 1. 3% 2. 4% Interjection 0. 9% 0% Euphemism 0. 3% 0. 8% use English names to address others (32. 9% vs. 18. 9%) men, on the other hand, more frequently use English words to reduce effort (22. 9% vs. 14. 8%). V. CONCLUSIONS We have described the construction of a corpus of Cantonese-English mixed code, based on speech transcribed from television programs in Hong Kong.Drawn from more than 60 hours of speech, this corpus is among the largest of its type. A novel feature of the corpus is the annotation of the motivation behind each code-mixed utterance. Having proposed a classification system for these motivations, we applied it on our corpus, and reported differences in the use of mixed code between genres, genders and age groups. A key finding is that four main motivations register, personal name, principle of economy, and specificity account for more than 95% of the embedded English segments.ACKNOWLEDGMENT This project was partially funded by a Small-Scale investigate Grant from the Department of Chinese, Translation and Linguistics at City University of Hong Kong. We thank Man Chong Mak and Hiu Yan Wong for compiling the corpus and performing annotation. REFERENCES 1 K. H. Y. Chen, The Social Distinctiveness of Two Code-mixing Styles in Hong Kong, in Proceedings of the 4th International Symposium on Bilingualism, MA Cascadilla Press, 2005, pp. 527541. J. Gumperz, The sociolinguistic significance of conversational code-switching, in RELC Journal 8(2), 1977, pp. 134. J.Gibbons, Code-mixing and koineizing in the speech of students at the university of Hong Kong, in Anthropological Linguistics 21(3), 1979, pp. 113123. B. H. -S. Chan, How does Cantonese-English code-mixing work? , in Language in Hong Kong at Centurys End, M. C. Pennington (ed. ), 1998, pp. 191216, Hong Kong Hong Kong University Press. D. C. S. Li, Linguistic convergence Impact of English on Hong Kong Cantonese, in Asian Englishes 2(1), 1999, pp. 536. K. K. Luke, Why two languages might be better than one motivations of language mixing in Hong Kong, in Language in Hong Kong at Centurys End, M.C. Pennington (ed. ), 1998, pp. 145159, Hong Kong Hong Kong University Press. D. C. S. Li, Cantonese-English code-switching research in Hong Kong a Y2K review, in World Englishes 19(3), 2000, pp. 305 322. H. Cao, teaching of a Cantonese-English code-mixing speech recognition system, PhD dissertation, Chinese University of Hong Kong, 2011. R. Appel and P. Muysken, Language contact and bilingualism. London Arnold, 1987. 2 3 4 5 6 Table 6 Distribution of code-switching motivations, contrasted between genders.Motivation Female Male Register 37. 5% 40. 7% Personal Name 32. 9% 18. 9% Principle of economy 14. 8% 22. 9% Specificity 10. 9% 13. 2% Quotation 1. 9% 1. 7% Doubling 1. 1% 1. 3% Interjection 0. 7% 1. 1% Euphemism 0. 3% 0. 2% Genders. Finally, we investigate whether codeswitching motivations are sloped according to gender. Aggregating statistics from both the current-affairs a nd talk shows, Table 6 compares the motivations of males and those of females. Females are shown to be more likely to 3 7 8 9 The speakers in the talk show are predominantly adults. 168