DEMoS: an Italian emotional speech corpus

Abstract

We present DEMoS (Database of Elicited Mood in Speech), a new, large database with Italian emotional speech: 68 speakers, some 9 k speech samples. As Italian is under-represented in speech emotion research, for a comparison with the state-of-theart, we model the 'big 6 emotions' and guilt. Besides making available this database for research, our contribution is three-fold: First, we employ a variety of mood induction procedures, whose combinations are especially tailored for specific emotions. Second, we use combinations of selection procedures such as an alexithymia test and self-and external assessment, obtaining 1,5 k (proto-) typical samples; these were used in a perception test (86 native Italian subjects, categorical identification and dimensional rating). Third, machine learning techniques-based on standardised brute-forced openSMILE ComParE features and support vector machine classifiers-were applied to assess how emotional typicality and sample size might impact machine learning efficiency. Our results are three-fold as well: First, we show that appropriate induction techniques ensure the collection of valid samples, whereas the type of self-assessment employed turned out not to be a meaningful measurement. Second, emotional typicality-which shows up in an acoustic analysis of prosodic main features-in contrast to sample size is not an essential feature for successfully training machine learning models. Third, the perceptual findings demonstrate that the confusion patterns mostly relate to cultural rules and to ambiguous emotions.

Machine Learning-Based Student Emotion Recognition for Business English Class

“…According to this theory, surprised and excited correspond to positive emotions; while anger, sad, hate, and fear correspond to negative emotions [18,19]. Positive emotions promote the occurrence of cognition activities, while negative emotions hinder the cognition process [20,21].…”

Corpora compilation for prosody-informed speech processing

“…The need for monolingual spoken data is growing steadily to achieve linguistic coverage in automatic speech recognition and text-to-speech research and development. Some examples to these are: the Switchboard corpus (Godfrey & Holliman, 1993), for English telephone conversational speech, the CALLHOME speech corpora, consisting of telephone conversations in several languages (Canavan, Graff, & Zipperlen, 1997), English Boston University Radio Speech Corpus (Ostendorf, Price, & Shattuck-Hufnagel, 1996), Rhapsodie (Lacheret et al, 2014), a French speech corpus with prosodic, syntactic and orthographic annotations, DEMoS (Parada-Cabaleiro et al, 2019) an Italian emotional speech corpus, RSC 3 https://rosettaproject.org/projects/300-languages/. (Georgescu et al, 2020), a Romanian read speech corpus for automatic speech recognition, TV3Parla (Külebi & Ö ktem, 2018) and ParlamentParla (Külebi et al, 2020), parliamentary and television speech corpora for Catalan.…”

Generating and Protecting Against Adversarial Attacks for Deep Speech-Based Emotion Recognition Models

“…For this study, we utilise the Database of Elicited Mood in Speech (DEMoS) [29], which is an Italian emotional speech corpus. DEMoS was collected from 68 speakers (23 females, 45 males) with 9 365 emotional and 332 neutral speech samples in total.…”

“…The 9 365 speech samples are annotated with seven classes of emotion shown in Table 1, of which all are used in our experiments. The emotions of DEMoS were induced by an arousal-valence progression [29]. To avoid speaker dependency during training, partitioning of the data (train, development, and test) was made speaker-independently with consideration to gender and emotional class balancing.…”