Dialogue Act Classification
Definition
Dialogue act classification annotates each conversational utterance with a speech act label that captures its illocutionary function: what the speaker is trying to do with the message rather than just what it says. Common dialogue act taxonomies include SWDA (Switchboard Dialogue Act) with 42 labels and ISO 24617-2 with a hierarchical scheme. Key acts include: inform (stating a fact), request (asking for action), question (asking for information), acknowledge (confirming understanding), and instruct (directing behavior). Classification uses sequence models that consider both the utterance content and its conversational context.
Why It Matters
Dialogue act classification enables more contextually aware chatbot responses by understanding the pragmatic intent of each message, not just its literal content. Recognizing that 'Can you tell me your hours?' is a question-request (not just an inquiry about capability) allows the bot to provide the specific information requested rather than answering the surface 'can you' literally. It also powers dialogue management systems that need to track conversation state across multiple turns and determine appropriate response strategies.
How It Works
Dialogue act classifiers use the current utterance combined with recent conversation history as input to a transformer-based classifier. Context is critical—'Sure' can be an acknowledgment, an affirmation, or agreement depending on what preceded it. Models typically encode the last 3-5 turns as a concatenated sequence with special separator tokens between turns. Fine-tuning on annotated conversational corpora (SWDA, MRDA, DailyDialog) produces models that generalize across conversation types. Joint models combine dialogue act prediction with slot filling and intent classification.
Dialogue Act Classification — Utterance → Act Label
Act label taxonomy
Classification results
"Hi there, can you help me?"
Greeting + Request"My order number is 84521."
Inform"Is that the right address?"
Confirm"Thanks, goodbye!"
ClosingReal-World Example
A customer service bot uses dialogue act classification to distinguish between a customer asking a question ('What is your return policy?') and making a complaint ('Your return policy is terrible!'). Questions trigger the knowledge retrieval pipeline; complaints activate an empathy response followed by escalation detection. This classification layer improved customer satisfaction scores by 18% by ensuring complaints were handled with acknowledgment before information, rather than jumping directly to policy details.
Common Mistakes
- ✕Conflating dialogue acts with intents—an intent captures what the user wants to achieve; a dialogue act captures the communicative function of the utterance
- ✕Ignoring conversational context—utterances classified in isolation are often ambiguous without preceding turns
- ✕Using overly fine-grained taxonomies without enough labeled data—many taxonomies have dozens of rarely-used labels that the model never learns
Related Terms
Intent Detection
Intent detection classifies user messages into predefined categories representing the user's goal—such as 'check order status' or 'report a bug'—enabling chatbots to route queries to the appropriate responses or workflows.
Dialogue Management
Dialogue management is the component of a conversational AI system that tracks conversation state and decides what the bot should do next — ask a follow-up question, retrieve information, take an action, or hand off to a human. It is the 'brain' that orchestrates a coherent, goal-directed conversation across multiple turns.
Slot Filling
Slot filling is the dialogue management process of collecting all the required pieces of information (slots) needed to complete a task. The chatbot systematically asks for any missing slots — like date, time, or account number — until it has everything needed to fulfill the user's request.
Natural Language Understanding (NLU)
Natural Language Understanding (NLU) is the AI capability that interprets the meaning behind human text or speech — identifying what the user wants (intent) and extracting key details (entities). NLU is the 'comprehension' layer of a chatbot, translating raw input into structured information the system can act on.
Natural Language Processing (NLP)
Natural Language Processing (NLP) is the field of AI focused on enabling computers to understand, interpret, and generate human language—powering applications from chatbots and search engines to translation and sentiment analysis.
Ready to build your AI chatbot?
Put these concepts into practice with 99helpers — no code required.
Start free trial →