Valeria Vieira dos Santos
Author directory2026
Quando o "..." importa: análise comparada de modalidades de transcrição de hesitações em entrevistas do Corpus Roda Viva
Valeria Vieira dos Santos
Proceedings of the 17th Brazilian Symposium in Information and Human Language Technology
Valeria Vieira dos Santos
Proceedings of the 17th Brazilian Symposium in Information and Human Language Technology
Este trabalho investiga como as hesitações — pausas, prolongamentos, truncamentos, repetições e preenchimentos lexicais — são representadas em diferentes modalidades de transcrição de entrevistas jornalísticas do português brasileiro. A partir de um corpus piloto de 95 turnos de fala do programa Roda Viva (TV Cultura), comparamos uma transcrição fiel (com marcas hesitativas) com uma versão higienizada e com uma transcrição em protocolos conversacionais (Jefferson; GAT2). Os resultados mostram que cada modalidade trata sistematicamente as marcas hesitativas de forma distinta, revelando a necessidade de um protocolo de anotação explícito. Propomos uma tipologia funcional inicial em cinco categorias formais e discutimos implicações para a construção de corpora orais e tarefas de anotação em PLN.
Scripted Hesitation in a Financial Customer-Service Chatbot: A Brazilian Portuguese Case Study
Valeria Vieira dos Santos
Proceedings of the 11th Workshop on Financial Technology and Natural Language Processing
Valeria Vieira dos Santos
Proceedings of the 11th Workshop on Financial Technology and Natural Language Processing
Financial customer service is a high-stakes conversational domain in which debt negotiation, cancellations, and complaints require pragmatic resources beyond fluent text generation. We present a case study of linguistic hesitation in approximately 3,000 WhatsApp customer-service interaction records (approximately 89k tokens) from one Brazilian financial company, contrasting clients, human agents, and one commercial scripted chatbot. A rule-based pipeline adapted a Conversation Analysis typology to written chat and identified 20,938 pattern occurrences across four formal categories. Client messages account for 59% of detections, but unavailable per-role token totals preclude rate comparisons. Archived within-role distributions place 40% of agent detections in sensitive contexts and 50% of chatbot detections in technical contexts. The chatbot also relies on only 3–5 recurrent surface forms. We interpret these results as descriptive evidence that human and scripted hesitation serve different interactional profiles, not as a general claim about financial chatbots or LLM-based agents. The study illustrates both the value of pragmatic analysis for financial conversational systems and the limits of conclusions drawn from a single-company, single-annotator corpus.