Abstract Echo questions serve two pragmatic functions (recapitulatory and explicatory) and are subdivided into two types (yes-no echo question and wh-echo question) in verbal communication. Yet to date, most relevant studies have been conducted in European languages like English and Spanish. It remains unknown whether the different functions of echo questions can be conveyed via prosody in spoken Chinese. Additionally, no comparison was made on the diversified algorithmic models in predicting functions by the prosodity of Chinese echo questions, a novel linguistic cognition in nature. This motivated us to use different acoustic cues to predict different pragmatic functions of Chinese echo questions by virtue of acoustic experiment and data modeling. The results showed that for yes-no echo question, explicatory function exhibited higher pitch and intensity patterns than recapitulatory function whereas for wh-echo question, recapitulatory function demonstrated higher pitch and intensity patterns than explicatory function. With regard to data modeling, the algorithm Support Vector Machine (SVM) relative to Random Forest (RF) and Logistic Regression (LR) performed better when predicting different functions using prosodic cues in both yes-no and wh-echo questions. This study from a digitized perspective adds evidence to the cognition of echo questions' functions on a prosodic basis.
Introduction
This study, published in Frontiers in Psychology, explores the intersection of linguistics, acoustics, and machine learning to decode how vocal signals convey different meanings in Mandarin Chinese. Specifically, it investigates echo questions—instances where a speaker repeats a previous utterance to seek clarification or express surprise.
Problem Statement: Prosody in Tonal Languages
When a speaker repeats someone’s words back as a question (e.g., "You bought a cake?"), they are using an echo question. These serve two primary purposes:
- Recapitulatory: Seeking repetition because the original statement was not heard clearly.
- Explicatory: Seeking a deeper explanation due to surprise or confusion.
While well-studied in non-tonal languages like English, Mandarin Chinese presents a unique challenge. In Mandarin, pitch is primarily used to determine lexical meaning (tones). This research addresses whether prosody (the "melody" of speech) can still signal pragmatic functions despite these tonal constraints and identifies which computational models best classify these differences.
Methodology
The researchers employed a two-phase approach involving acoustic analysis and machine learning:
- Acoustic Analysis: 20 native Mandarin speakers recorded 720 sentences across four conditions: Yes-No vs. Wh-echo questions, each serving Recapitulatory vs. Explicatory functions.
- Feature Extraction: Nine prosodic parameters were measured, including mean, maximum, and minimum pitch (f0), duration, and intensity.
- Data Modeling: Three supervised learning algorithms—Random Forest (RF), Logistic Regression (LR), and Support Vector Machine (SVM)—were compared for their ability to classify question functions based on acoustic data using 10-fold cross-validation.
Key Insights: The Prosodic "Flip"
A central finding of the study is that the vocal signaling of pragmatic function depends heavily on the question type.
In Yes-No questions, the explicatory function is characterized by higher pitch and intensity ("louder and higher"). Conversely, in Wh-echo questions, the recapitulatory function receives the boost in pitch and intensity.
The authors suggest this "flip" occurs because Wh-words (like "what") are inherently interrogative. To signal a recapitulatory intent, speakers must apply extra vocal effort. In contrast, Yes-No questions often end in nouns that may sound declarative; thus, higher pitch is required to signal an explicatory request rather than simple confirmation.
Results and Predictive Modeling
The study demonstrated that machine learning can effectively distinguish subtle linguistic intentions:
- Predictive Accuracy: SVM outperformed other models in both categories. It achieved 83% accuracy for Yes-No questions and 86% accuracy for Wh-echo questions.
- Feature Significance: For Yes-No questions,
f0Maxandf0Rangewere the strongest indicators. For Wh-echo questions,f0MeanandintensityMaxwere the most significant features.
Conclusion and Future Directions
While the study provides strong evidence for "digitized cognition" in linguistics, limitations remain. The interaction between lexical tone and sentence prosody requires further control, and the sample size (20 participants) is relatively small for complex machine learning. Future work aims to integrate these models into AI and speech recognition systems to enhance real-time understanding of human intent and emotion.
Summary updated