Developments in Human-Computer Interaction and Natural Language Processing

The field of human-computer interaction and natural language processing is moving towards a more nuanced understanding of human language and behavior. Researchers are exploring the use of multimodal approaches, incorporating speech, text, and visual cues to improve communication between humans and machines. The development of new datasets and models is enabling the detection of subtle language phenomena, such as sarcasm and satire, and the classification of interjections. These advances have the potential to improve human-computer interaction, financial forecasting, and discourse evaluation. Noteworthy papers include: SeLeRoSa, which introduces a sentence-level Romanian satire detection dataset and evaluates baseline models. Beyond Words, which presents a novel task of interjection classification and publishes a dataset of interjection signals. Multimodal Proposal for an AI-Based Tool, which introduces a framework for generating semantically rich embeddings of earnings calls. Spoken in Jest, Detected in Earnest, which provides a systematic review of sarcasm recognition and highlights the importance of leveraging speech data for automatic sarcasm recognition.

Sources

SeLeRoSa: Sentence-Level Romanian Satire Detection Dataset

Beyond Words: Interjection Classification for Improved Human-Computer Interaction

Multimodal Proposal for an AI-Based Tool to Increase Cross-Assessment of Messages

Spoken in Jest, Detected in Earnest: A Systematic Review of Sarcasm Recognition -- Multimodal Fusion, Challenges, and Future Prospects

Built with on top of