Hesitations in Urdu/Hindi: Distribution and Properties of Fillers & Silences

Abstract

This research presents an analysis of hesitations in Urdu/Hindi semi-spontaneous dialogues. We annotated and analyzed twenty-five minutes of speech to investigate the frequency of hesitations and the properties of fillers as well as the formants in fillers' vocalic intervals to determine their vowel quality. We found that our participants used fillers, silences, and prolongations with varying frequency. Moreover, Urdu/Hindi speakers used the fillers with only vocalic intervals (uh) more frequently than the ones with vocalic intervals followed by nasals (um). The regression analysis showed that the um type fillers were significantly longer and followed by longer silences as compared with the uh type fillers. Furthermore, the um types were placed more frequently at the turn medial position, whereas the uh type fillers occurred at turn initial or medial position with similar frequency. The analysis of their formants showed that the vocalic intervals used in the fillers differed from other vowels in the inventory of Urdu/Hindi. Our data confirms the existing claim that uh and um are two distinct types of fillers. Our results are relevant for developing speech synthesis systems for Urdu/Hindi as well as improving the existing models seeking to incorporate hesitations and fillers in a realistic manner.

Cognitive Load Increases Spoken and Gestural Hesitation Frequency

“…A silence, as per our working definition stated above, and as conceived by Trouvain and Werner (2022), needs to be audible as well as visible in the signal, thus inserting a pause into the speech signal. A filler also stands out markedly from the surrounding speech, as its vowel quality and pitch contour is usually different from other vowels in the language system or from the surrounding speech, respectively (Belz 2021;Belz and Reichel 2015;Jabeen and Betz 2022). Lengthening, however, cannot be isolated from the flow of speech.…”

Variability in hesitations in Punjabi semi-spontaneous narrative speech: An automatic clustering based analysis

“…The type and distribution of hesitations has been reported to be language and speaker specific. For example, the FPs used by Japanese speakers [8] differ from those used by speakers of Hungarian [9], English [10], German [11], Urdu/Hindi [12], and European Portuguese [13]. The position and duration of SPs has also been shown to be variable [14,12].…”

“…For example, the FPs used by Japanese speakers [8] differ from those used by speakers of Hungarian [9], English [10], German [11], Urdu/Hindi [12], and European Portuguese [13]. The position and duration of SPs has also been shown to be variable [14,12]. Speaker based variation in the frequency and selection of hesitation types has been reported by [15,16].…”

“…As for FPs, the most frequently used filled pauses in English are vowel only or vowel-nasal sequences [10]. But FPs have been reported to consist of only a nasal consonant, a cluster of vowel-nasalfricative, as well as a sequence of two vowels in Hungarian [9] and Urdu/Hindi [12]. Recently, [11] showed that FPs in German may consist of glottal stops as well as bisyllabic forms.…”

“…To this day, most of the existing research on hesitations is concerned with the European languages, and South Asian languages have been largely neglected. To our knowledge, [12] is the only systematic investigation of hesitations in Urdu/Hindi, a South Asian language, while there is no analysis of hesitations in Punjabi so far. In the current study, we aim to fill this gap and analyze the distribution and properties of hesitations in Punjabi semi-spontaneous narrative speech.…”