5. Que el director estigui acompanyat sempre que es facin entrevistes
8.6. Pla d'acció per a la comercialització
In this section, the word and sentence tests that were used to assess speech intelligibility in the first phase of the study are described. Three types of tests were used. The first word test was the Diagnostic Rhyme Test (DRT) and the second word test was the Phonemically Balanced word test (PB word). Additionally, Phonemically Balanced sentence lists (PB sentence) were used in order to compare the word and sentence test scores, and to identify any possible effects of context on speech intelligibility.
3.3.1.1 Diagnostic Rhyme Test (DRT)
The DRT is a listening test consisting of 192 words arranged in 96 pairs (ANSI/ASA.S3.2, 2009). The words are common, monosyllabic words, and most of them have three sounds ordered in a consonant-vowel-consonant (CVC) sequence. The word pairs differ only in their initial consonants, therefore discrimination of each distinctive feature of a given language can be analysed in relation to the room acoustic properties. The DRT is a closed-set type of test, which requires participants to select the
(a) An example list of the English DRT (Voiers, 1977).
(b) An example list of the Arabic DRT (Boudraa et al, 2008).
Tenseness Nasality Flatness
(c) An example list of the Mandarin DRT (Fu et al., 2011).
Airflow Nasality Sustention
correct word out of the provided pair of words. No carrier sentence is needed to conduct the DRT test (ANSI/ASA.S3.2, 2009).
Jakobson et al. (1952) suggested that the sounds of languages can be identified by using a set of distinctive features, which does not exceed twelve distinctive features. The test focuses on consonants in order to compare distinctive features between two sounds. The English DRT test focuses on voicing, nasality, sustention, sibilation, graveness, and compactness (ANSI/ASA.S3.2, 2009); the Arabic DRT test focuses on tenseness, nasality, mellowness, flatness, graveness, and compactness (Boudraa et al., 2008), and the Mandarin DRT test focuses on airflow, nasality, sustention, sibilation, graveness, and compactness (Fu et al., 2011). Full lists and detailed information on distinctive
language is presented in Table 3.2 (a), Table 3.2 (b), and Table 3.2 (c). All the DRT material is available in Appendix A.
The lack of DRT material in Polish is a limitation of the current study, although comparisons between DRT and PB English words data (the former being taken from the current study and the latter from (Anderson and Kalb, 1987)) indicate that the variability between DRT and PB scores tends to be fairly small (Figure 3.1), suggesting that comparisons between DRT and PB results are acceptable (as also pointed out in ANSI/ASA S3.2-2009 (ANSI/ASA.S3.2, 2009)). Figure 3.1 shows that DRT and PB scores of English have an average difference (calculated from absolute values) of 2.4%
across the four STI conditions considered, with a maximum difference of 5.5%
observed at STI = 0.4. This is well below the large differences observed between languages that are presented in section 3 (which are as high as 33% at STI = 0.4), indicating that these inaccuracies are not expected to have affected the main findings obtained when comparing Polish PB word scores to DRT scores of the other languages.
It is however accepted that some inaccuracies should be expected and are unfortunately not quantifiable for Polish, and that the variations between DRT and PB word scores of Polish could be higher than those presented for English in Figure 3.1. The data taken from (Anderson and Kalb, 1987) was based on the standard Harvard PB word test, which is commonly used in the United States. It should also be noted that comparability of DRT and PB scores can be achieved only by removing the effect of guesswork in the calculation of DRT scores (see equation (4.1)), as rhyme tests are closed tests that are otherwise expected to provide higher scores (Peng, 2011).
Table 3.3 An example list of the Polish PB words (Ozimek et al., 2007).
Dres Jas Jacht Mech Biust Tak
Glos Wyz Pyl Plaszcz Won Krzak
Zal Kant Krem Wodz Klops Bron
Figure 3.1 DRT (present study) and PB (Anderson and Kalb, 1987) word intelligibility scores of English obtained under different STI conditions (data markers, standard errors
of the means (for DRT data only) and logarithmic regression lines).
3.3.1.2 Phonemically balanced word test (PB word)
A listener of a spoken language has the knowledge of linguistic information, such as, phonological, lexical, syntactic, and semantic properties of that language (Kalikow et al., 1977). Improved intelligibility of such linguistic properties provide a context, eventually decreasing the dependency on the perfectly transmitted acoustical signal.
Therefore, a speech intelligibility test that demonstrates an everyday speech must assess both the acoustic-phonetic and the linguistic-structural components of the process (Kalikow et al., 1977). In order to achieve this tests, similar to the phonemically balanced word tests (PB word) were created in order to represent linguistic properties of languages.
PB word lists represent a language by having approximately the same phonetic properties and sound types of that language. It should be noted that in order to represent a specific language, all of the word lists must be phonemically balanced. Therefore, the DRT is a phonemically balanced test as well. The difference between the DRT and PB word tests is that the former focuses on the differentiation of initial consonants, and the latter focuses on the intelligibility of the whole word (ANSI/ASA.S3.2, 2009). The Polish PB word lists used in the current study consists of 4 sets of 48 words, with one
0
presented in Table 3.3. All the Polish PB words are available in Appendix B.
3.3.1.3 Phonemically balanced sentence test (PB sentence)
In the current study, the effects of context on speech intelligibility were also investigated. Typically, intelligibility of a given word increases when it is presented in a sentence, compared to an isolated word, when context of a sentence is adequately transferred to the listener (Miller et al., 1951). The context of a sentence is dependent on the number of possible alternative words that are available for a given sentence in that particular word location. More alternatives lead to a less predictable sentence, and less alternatives lead to a higher degree of predictability (Kalikow et al., 1977). It should be noted that levels of predictability were not taken into account for the statistical analysis of the research presented, due to the fact that the PB sentence lists were not designed accordingly (except the English PB sentence test) and the scores were calculated by counting the number of correct words in the whole sentence, rather than evaluating only the last word. The PB sentence lists that were used in the current study were composed of everyday sentences, therefore representing a wide range of predictability. The English PB sentence pool consisted of 6 high predictability (HP) and 4 low predictability (LP) sentences. Polish, Arabic, and Mandarin sentences pools consisted of sentences that represent an everyday conversation. The number of words in each sentence varied with the language. In the English PB sentence list, there were a minimum of 6 and a maximum of 7 words in each sentence. In the Arabic PB sentences the minimum number of words was 3 and the maximum was 6. In the Mandarin list, all of the sentences had 7 words. Lastly, in the Polish PB list the number of words was 5. The PB sentence lists consisted of a total of 10 sentences. 2 sentences were used for STI = 0.8 and STI = 0.6 conditions, and 3 sentences for STI = 0.4 and STI = 0.2 were selected randomly from the sentences pool (Kalikow et al., 1977) (Ozimek, 2009) (Boudraa et al., 2000) (Fu et al., 2011). An example set of sentences for each language is presented in Table 3.4 (a) – (d). All the sentences used are available in Appendix C.
(a) An example list of the English PB sentences (Kalikow et al., 1977).
The watchdog gave a warning growl HP She made the bed with clean sheets HP The old man discussed the dive LP
(b) An example list of the Polish PB sentences (Ozimek, 2009).
Znowu ta winda nie dziala Najpierw zwabilo go swiatlo Wracam pozno do hotelu
(c) An example list of the Arabic PB sentences (Boudraa et al., 2000).
(d) An example list of the Mandarin PB sentences (Fu et al., 2011).
The recording procedure of the word and sentence lists, specifications of the speakers and the equipment that was used are presented in the next section.