Abstract. This paper performs a global analysis of entrainment between dyads in map-task dialogues in European Portuguese (EP), including 48 dialogues, between 24 speakers. Our main goals focus on the acoustic-prosodic similarities between speakers, namely if there are global entrainment cues displayed in the dialogues, if there are degrees of entrainment manifested in distinct sets of features shared amongst the speakers, if entrainment depends on the role of the speaker as either giver or follower, and also if speakers tend to entrain more with specific pairs regardless of the role. Results show global entrainment in almost all the dyads, but the degrees of entrainment (stronger within the same gender), and the role effects tend to be less striking than the interlocutors' effect. Globally, speakers tend to be more similar to their own speech in other dialogues than to their partners. However, speakers are also more similar to their interlocutors than to speakers with whom they never spoke.

Automatic personality analysis has gained attention in the last years as a fundamental dimension in human-to-human and human-to-machine interaction. However, it still suffers from limited number and size of speech corpora for specific domains, such as the assessment of children's personality. This paper investigates a semi-supervised training approach to tackle this scenario. We devise an experimental setup with age and language mismatch and two training sets: a small labeled training set from the Interspeech 2012 Personality Sub-challenge, containing French adult speech labeled with personality OCEAN traits, and a large unlabeled training set of Portuguese children's speech. As test set, a corpus of Portuguese children's speech labeled with OCEAN traits is used. Based on this setting, we investigate a weak supervision approach that iteratively refines an initial model trained with the labeled data-set using the unlabeled data-set. We also investigate knowledge-based features, which leverage expert knowledge in acoustic-prosodic cues and thus need no extra data. Results show that, despite the large mismatch imposed by language and age differences, it is possible to attain improvements with these techniques, pointing both to the benefits of using a weak supervision and expert-based acoustic-prosodic features across age and language.

This paper investigates the correlation between the prosodic properties and pragmatic functions of affirmative constituents in adult-adult interactions in European Portuguese (CORAL corpus). 515 affirmative constituents produced in 460 answers, extracted from 11 dialogues between 12 speakers, were analyzed. Results show that: i) sim 'yes', ok and grunts are the most frequent affirmative constituents; ii) sim 'yes' is associated with all the communicative functions analyzed, agreement, auto positive and confirm, ok tends to occur with agreement, and grunts are mainly associated to auto positive; iii) affirmative constituents have different prosodic properties according to their pragmatic function: agreement and confirm show a similar behavior, being auto positive the most distinct function. Agreement and confirm are commonly uttered with (H+)L* L%, whereas auto positive is commonly uttered with L*+H / (L+)H* H%. When affirmative constituents co-occur in the same answer, there are evidences of tone copying between them. Correlations between constituents were also found in the following parameters: energy, pitch mean, maxima and minima, as well as pitch range. As for context-answer pairs, a pitch concord effect is also found between the pairs instruct-agreement and propositional question-confirm, although expressed in different degrees.

This paper presents an analysis of discourse markers in two spontaneous speech corpora for European Portuguese - university lectures and map-task dialogues - and also in a collection of tweets, aiming at contributing to their categorization, scarcely existent for European Portuguese. Our results show that the selection of discourse markers is domain and speaker dependent. We also found that the most frequent discourse markers are similar in all three corpora, despite tweets containing discourse markers not found in the other two corpora. In this multidisciplinary study, comprising both a linguistic perspective and a computational approach, discourse markers are also automatically discriminated from other structural metadata events, namely sentence-like units and disfluencies. Our results show that discourse markers and disfluencies tend to co-occur in the dialogue corpus, but have a complementary distribution in the university lectures. We used three acoustic-prosodic feature sets and machine learning to automatically distinguish between discourse markers, disfluencies and sentence-like units. Our in-domain experiments achieved an accuracy of about 87% in university lectures and 84% in dialogues, in line with our previous results. The eGeMAPS features, commonly used for other paralinguistic tasks, achieved a considerable performance on our data, especially considering the small size of the feature set. Our results suggest that turn-initial discourse markers are usually easier to classify than disfluencies, a result also previously reported in the literature. We conducted a cross-domain evaluation in order to evaluate the robustness of the models across domains. The results achieved are about 11%-12% lower, but we conclude that data from one domain can still be used to classify the same events in the other. Overall, despite the complexity of this task, these are very encouraging state-of-the-art results. Ultimately, using exclusively acoustic-prosodic cues, discourse markers can be fairly discriminated from disfluencies and SUs. In order to better understand the contribution of each feature, we have also reported the impact of the features in both the dialogues and the university lectures. Pitch features are the most relevant ones for the distinction between discourse markers and disfluencies, namely pitch slopes. These features are in line with the wide pitch range of discourse markers, in a continuum from a very compressed pitch range to a very wide one, expressed by total deaccented material or H+L* L* contours, with upstep H tones.

This work describes the discourse markers present in two corpora for European Portuguese, in different domains (university lectures and map-task dialogues). In this study, we also perform a multiclass automatic classification task based on prosodic features to verify in both corpora which words are discourse markers, which are disfluencies, and which are sentence like-units (SUs). Results show that the selection of discourse markers varies across domain and between speakers. As for the classification task, results show that the discourse markers are better classified in the lectures corpus (87%) than in the dialogue corpus (84%). However, cross-domain experiments evidenced that data trained with the dialogue corpus predicts better the events in the lecture corpus, since this domain displays more speakers and therefore complex patterns. In both corpora, markers are more easily classified as SUs than as disfluencies.