Can We Detect Irony in Speech Using Phonetic Characteristics Only? – Looking for a Methodology of Analysis

Результат исследований: Публикации в книгах, отчётах, сборниках, трудах конференцийстатья в сборнике материалов конференции

Аннотация

The current paper aims to investigate the perception of Russian ironic laboratory speech and to develop a methodological basis for the further research on perceptually relevant acoustic cues of irony. On the first stage of analysis we created short monologues and dialogues including ironic and homonymous non-ironic sentences. Two Russian native male speakers read this material. The recording was accomplished in a sound booth. Then, randomized fragments of the recordings were suggested to native listeners in an auditory perception experiment. The mean intensity, duration of the stressed vowel and F0 range in the reliably identified target fragments were analyzed, as well as functional intonation models in terms of the system suggested by E. Bryzgunova. The results showed that the listeners are able to detect irony without any contextual or lexical marker of it. The acoustic analysis of the reliably identified stimuli demonstrated the difference in intensity level and stressed vowel duration between ironic and non-ironic utterances. The obtained results allow creating a methodology of data collection and validation for the construction of ironic speech corpus and its analysis. They may also have applications in teaching Russian as a second language, human-machine communication and improving natural language processing systems.

Язык оригиналаанглийский
Название основной публикацииSpeech and Computer - 22nd International Conference, SPECOM 2020, Proceedings
РедакторыAlexey Karpov, Rodmonga Potapova
ИздательSpringer Nature
Страницы544-553
Число страниц10
ISBN (печатное издание)9783030602758
DOI
СостояниеОпубликовано - 2020
Событие22nd International Conference on Speech and Computer, SPECOM 2020 - St. Petersburg, Российская Федерация
Продолжительность: 7 окт 20209 окт 2020

Серия публикаций

НазваниеLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Том12335 LNAI
ISSN (печатное издание)0302-9743
ISSN (электронное издание)1611-3349

конференция

конференция22nd International Conference on Speech and Computer, SPECOM 2020
СтранаРоссийская Федерация
ГородSt. Petersburg
Период7/10/209/10/20

Предметные области Scopus

  • Теоретические компьютерные науки
  • Компьютерные науки (все)

Fingerprint Подробные сведения о темах исследования «Can We Detect Irony in Speech Using Phonetic Characteristics Only? – Looking for a Methodology of Analysis». Вместе они формируют уникальный семантический отпечаток (fingerprint).

Цитировать