General introduction

This research group operates with the support and funding of the Hungarian National Research, Development and Innovation Office of Hungary, project No. NKFIH-K-128810 called The interrelations of phonetic properties in speech units. The aim of this research is to examine phonetic parameters and their interrelations in real communication situations, combining phonetic, psycholinguistic and conversational-analysis aspects, along with formal and functional approaches. The material of the research mainly consists of recordings from the Hungarian Spontaneous Speech Database. Using multi-participant speech recordings is necessary because it provides an opportunity to analyse speech units, which have been phonetically less analysed, such as turns, turn-taking or overlapping speech.

Multi-aspect analysis of discourse markers as verbal discourse-organizing units is another important research aim. Large data sample gives the researcher the ability to describe ongoing changes such as formation of discourse marker combinations and realisation of opinion marker expressions, as well as their interrelation with suprasegmental features. An acoustic analysis of laughter as a nonverbal vocal element is intended to be carried out by combining pragmatic and sociolinguistics aspects. Furthermore, a high priority aim of this research is a rich description of speech accommodation, which has been scarcely investigated in relation with the suprasegmental structure of speech, thus, it is a novelty in Hungarian speech research.

The research has further scientific benefits as a result of the cooperation of several linguistic disciplines, that the system of conversations might be described. If discourses are performed according to a particular system (which not only corresponds to discourse organizers but also to some phonetic patterns), theoretically they can be modelled technologically. Thus, in addition to their scientific merits, the results might be used in speech technology applications like the description of laughter and its modelling in speech technology.

About the database

The BEA is a database of approximately 500 records. Each recording contains a total of 8 subtasks of which, the 'conversation' aspect was selected for the present study. The recordings were quasi-spontaneous as the topic of conversation had been added at the beginning of the recording, without any pre-fixed protocol or script to be followed. The participants had no time to plan the speech production, and the outcome depended on the opinion and utterances of others. The conversation part is the sixth subtask of the recording. Each conversation had three participants. The Interviewer (I) and the Conversational Partner (CP) were the same people on all recordings: two monolingual, Hungarian female colleagues of the same age, both from Budapest and awarded with the same scholarly degree. They were 28 when the interviews started and are now 35 (the interviews and building the database took almost 10 years to complete). The Subjects (S) were different people in each conversation. For our study, 5 male and 5 female monolingual Hungarians from Budapest were selected, aged 20-35. The third participant, a colleague of the interviewer, yet unknown to the speaker, was invited to the studio to manage the recordings. The participants talked about an everyday topic, prepared by the interviewer in advance. The recordings were not pre-constructed, quasi-spontaneous ones, but the interviewer tried to maintain spontaneous conversations with the participants for as long as possible.

The existing six-tier annotation (speech section, word level for 3 speakers) was complemented with further tiers by the software Praat (Boersma & Weenink, 2018): (i) backchannel responses; (ii) speaking at the same time; (iii) turns; (iv) turn-takings; (v) speaker changes; (vi) whether a given turn belongs to a narrative or a dialogic (differentiated by text structure organization, cf. Tolcsvai, 2001; Hutchby & Woofitt, 2006) section inside the global structure (see Figure 1). The whole dataset is 131 minutes long (mean length: 13 min, range: 6.9-23.3 minutes).

Figure 1. Example annotation (to make the figure more comprehensible, the word level was omitted; Meaning of the inscriptions: 'IK szakasz' = the I's IPUs + silent pauses, 'B szakasz' = S's IPUs + silent pauses, 'TP szakasz' = CP's IPUs + silent pauses, 'együttbeszélés' = overlapping speech, 'váltás' = turning points, 'forduló' = the main speaker of the current section, 'narr/dial' = narrative or dialogic speech section; 'AHA' = humming, 'ő egy hűtőházban dolgozik' = he works in a refrigeratory, 'milyen munka ez hogy' = what kind of work is that?,'JA' = okay)

The research was supported by the Hungarian National Research, Development and Innovation Office of Hungary, project No. NKFIH-K-128810. The research group works in the Phonetic Department of the Hungarian Research Centre for Linguistics.