Partial Utterance Detection for Virtual Agents
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spoken language understanding systems struggle to handle partial or non-sentential utterances, which lack sufficient information for effective interpretation and response, leading to difficulties in determining user intent and maintaining natural-sounding conversations.
Innovation Solution
A system and method for detecting and classifying partial utterances, using sets of rules to identify partial utterances and providing advice codes for handling them, allowing the system to bypass traditional intent determination and slot naming processes, and recommending responses or actions to advance the conversation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a spoken language understanding system uses traditional intent determination and slot naming processes, then it can handle complete sentences effectively, but it fails to accurately interpret partial utterances which lack sufficient information
Solution Approach 1:
The system segments the intent determination process into two distinct paths: a traditional path for complete sentences and a specialized path for partial utterances. The detection module divides utterances into categories (complete, partial, or neither) and routes them through different processing sequences, allowing each path to be optimized for its specific input type without compromising the other.
Solution Approach 2:
A detection module acts as an intermediary between the utterance input and the intent determination system. This intermediary detects whether an utterance is complete or partial and provides routing information to guide the subsequent processing, enabling the system to adapt its interpretation strategy based on the utterance type without requiring complete restructuring of the overall architecture.
2Productivity
If the system processes all utterances through the traditional spoken language understanding pipeline, then it maintains a consistent processing framework, but it wastes time and resources on partial utterances that cannot be properly interpreted
Solution Approach 1:
The system performs preliminary detection of utterance completeness before initiating the full intent determination process. By detecting whether an utterance is partial in advance, the system can prevent unnecessary processing of uninterpretable inputs, saving computational resources and time while maintaining efficient processing of valid complete utterances.
Solution Approach 2:
For partial utterances, the system performs only the detection and routing actions necessary to identify them as such, rather than executing the complete intent determination pipeline. This partial processing approach avoids the excessive action of attempting full interpretation on inputs that cannot be properly understood, thereby improving overall processing efficiency.
3Reliability
If the system attempts to determine user intent from partial utterances using traditional methods, then it maintains a unified processing approach, but it produces inaccurate or meaningless results due to insufficient information
Solution Approach 1:
The processing framework is segmented into distinct pathways: one for complete utterances that follows the traditional intent determination process, and another for partial utterances that is routed to alternative handling mechanisms. This segmentation ensures that each pathway is appropriate for its input type, maintaining result accuracy while managing complexity through structured organization.
Solution Approach 2:
Instead of attempting to force partial utterances through the traditional intent determination pipeline and risking inaccurate results, the system inverts the approach by detecting partial utterances first and then routing them to alternative processing methods. This inversion prevents the generation of meaningless results by avoiding inappropriate processing of insufficient inputs.
Data Source
AI summary
The system and method generally include identifying whether an utterance spoken by a user (e.g., customer) is a complete or incomplete sentence. For example, the system may include a partial utterance detection module that determines whether an utterance spoken by a user is a partial utterance. The detection process may include providing a detection advice code that gives a recommendation for handling the utterance of interest. If it is determined that the utterance is an incomplete sentence, then the system and method can identify the type of utterance. For example, the system may include a partial utterance classification module that predicts the class of a partial utterance. The classification process may include providing a classification advice code that gives a recommendation for handling the utterance of interest. Once a partial utterance is detected and classified, the system and method can further determine what the user meant by the utterance and can recommend a response to the user's utterance that further advances a conversation between the user and a virtual agent.