ProsoTool, a method for automatic annotation of fundamental frequency

Szekrényes, István

doi:10.1109/coginfocom.2015.7390606

Cited by 7 publications

(4 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In Figure 3, the final output can be seen with the segmented and labeled F0 plots using five possible categories of intonation contour (rise, fall, ascending, descending and level ), depending on the amplitude and the duration of the modulation. Further details on the method are described in [18]. The same algorithm is available at the official website 4 of the e-magyar project under the name of emPros (in accordance with the naming guideline of the project).…”

Section: Annotation Of Intonationmentioning

confidence: 99%

Classification of Formal and Informal Dialogues Based on Turn-Taking and Intonation Using Deep Neural Networks

Szekrényes

Kovács

2017

Speech and Computer

View full text Add to dashboard Cite

Abstract. Here, we introduce a classification method for distinguishing between formal and informal dialogues using feature sets based on prosodic data. One such feature set is the raw fundamental frequency values paired with speaker information (i.e. turn-taking). The other feature set we examine is the prosodic labels extracted from the raw F0 values via the ProsoTool algorithm, which is also complemented by turn-taking. We evaluated the two feature sets by comparing the accuracy scores our classification method got, which uses them to classify dialogue-excerpts taken from the HuComTech corpus. With the ProsoTool features we achieved an average accuracy score of 85.2%, which meant a relative error rate reduction of 24% compared to the accuracy scores attained using F0 features. Regardless of the feature set applied, however, our method yields better accuracy scores than those got by human listeners, who only managed to distinguish between formal and informal dialogue to an accuracy level of 56.5%.

show abstract

Section: Annotation Of Intonationmentioning

confidence: 99%

Classification of Formal and Informal Dialogues Based on Turn-Taking and Intonation Using Deep Neural Networks

Szekrényes

Kovács

2017

Speech and Computer

View full text Add to dashboard Cite

show abstract

“…Audio was annotated for the classes of intonation phrase, emotions and discourse, in addition to phonetic events within speech, such as silence, hesitation, restart, non-phonemic sounds, and noise. Automatic methods were applied to the annotation of the phonetic features of the sound track: in addition to marking the absolute values for F0 and intensity, a special algorithm (Szekrényes, 2014, 2015) was used to annotate stylized intonation and intensity contours of speech in order to capture the contribution of speech prosody to the multimodal expression of the pragmatic content of the interaction. The pragmatic levels of annotation included the classes of turn management, attention, agreement, deixis and information structure.…”

Section: The Hucomtech Corpus: Its Structure and Annotation Schemementioning

confidence: 99%

Agreeing/Disagreeing in a Dialogue: Multimodal Patterns of Its Expression

Hunyadi

2019

Front. Psychol.

View full text Add to dashboard Cite

Multimodality is a key property of communication. Even though the main channel of exchanging information is textual, the written text also relies on paralinguistic means to convey additional information using various kinds of punctuation. Speech, too, is always present in contributing to the understanding of the wider scope of the context, represented by some restricted means, including typography, typesetting, and style. Gestures are also part of the text in its broader sense: public stage performances necessarily accompany the text, too. The complex of text, speech and gestures can be understood as a set of unified multimodal patterns that progress in time. This paper offers a comprehensive insight in the temporal structure of dialogue interaction through the study of the multimodal expression of agreement/disagreement in the HuComTech Corpus, presenting patterns of behavior discovered using the research environment Theme .

show abstract

“…For silences, the same representation was used as in the previous experiment. Segments of intonational events (the term was adapted from Taylor, 2000) were generated by a rule-based algorithm (Szekrényes, 2015) which was developed for automatic annotation of stylized shapes of intonation. This labeling method was applied to the speech of every related actor ( agentf 0 mov , speakerf 0 mov , overlapf 0 mov - overlapping speech was considered as a “third speaker” here) using five possible categories: rise , fall , level , ascending and descending .…”

Section: Research Materialsmentioning

confidence: 99%

Post-processing T-patterns Using External Tools From a Mixed Method Perspective

Szekrényes

2019

Front. Psychol.

View full text Add to dashboard Cite

Researchers of behavioral science often work with time-aligned annotation data based on video and/or audio recordings. Various platforms are available to process these data, offering various kinds and ways of data analysis. It often happens that one would wish to use one platform for a certain kind of analysis, and another platform for another kind. It may also happen that one would keep the results of the first analysis and continue processing the data using another platform—all as a chain of analyses on the way to discovery. When it comes to T-pattern analysis, the task of further analyzing already identified patterns across platforms requires a general framework within which virtually any kind of data can be processed in a cross platform environment: that of a database. Data (including patterns) from one platform are then imported into this database, where these patterns are further processed to uncover further properties, then the patterns can be exported to another platform, including the one the data originated from. This contribution aims at introducing a new methodology and a tool implemented as a web-based service for these purposes.

show abstract

ProsoTool, a method for automatic annotation of fundamental frequency

Cited by 7 publications

References 10 publications

Classification of Formal and Informal Dialogues Based on Turn-Taking and Intonation Using Deep Neural Networks

Classification of Formal and Informal Dialogues Based on Turn-Taking and Intonation Using Deep Neural Networks

Agreeing/Disagreeing in a Dialogue: Multimodal Patterns of Its Expression

Post-processing T-patterns Using External Tools From a Mixed Method Perspective

Contact Info

Product

Resources

About