Hierarchical Parse-Based Dialog Model for Spoken Language Understanding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current dialog modeling approaches underutilize or ignore annotations in spoken dialog corpora, limiting their ability to understand and reason about dialog structure effectively.
Innovation Solution
The development of systems and methods for dialog modeling that receive annotated spoken dialogs, parse them using a hierarchical, parse-based dialog model operating incrementally from left to right, and construct a functional task structure to interpret user utterances, plan system responses, and predict next user utterances, employing shift-reduce, start-complete, or connection path models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If annotations in spoken dialog corpora are underutilized or ignored, then the dialog modeling approach is simpler, but the ability to understand and reason about dialog structure is limited
Solution Approach 1:
The patent applies preliminary action by pre-annotating spoken dialog corpora with dialog acts and task/subtask information before modeling. This pre-processing step prepares the data in advance, allowing the model to leverage rich structural clues during training and inference, thereby improving dialog structure understanding without adding complexity to the core modeling algorithm
Solution Approach 2:
The patent introduces an intermediary layer of annotations (dialog acts and task/subtask labels) that mediate between raw spoken dialogs and the dialog model. These annotations serve as intermediate representations that capture essential dialog structure, enabling the model to reason about dialog flows more effectively while keeping the model architecture itself relatively simple
2Measurement precision
If a hierarchical parse-based dialog model is used to parse spoken dialogs incrementally, then the dialog structure is better captured, but the processing time increases
Solution Approach 1:
The patent applies segmentation by dividing the dialog parsing task into incremental steps that process dialogs from left to right. The hierarchical parse-based model breaks down complex dialog structures into smaller parseable units, allowing incremental construction of the dialog representation without requiring complete re-processing of entire dialogs
Solution Approach 2:
The model uses preliminary action by pre-defining hierarchical parse structures and rules that guide the incremental parsing process. This preparation allows the system to efficiently build dialog representations step-by-step without exploring all possible parse configurations at each stage, reducing overall processing time while maintaining structural accuracy
Data Source
AI summary
Disclosed herein are systems, computer-implemented methods, and computer-readable media for dialog modeling. The method includes receiving spoken dialogs annotated to indicate dialog acts and task/subtask information, parsing the spoken dialogs with a hierarchical, parse-based dialog model which operates incrementally from left to right and which only analyzes a preceding dialog context to generate parsed spoken dialogs, and constructing a functional task structure of the parsed spoken dialogs. The method can further either interpret user utterances with the functional task structure of the parsed spoken dialogs or plan system responses to user utterances with the functional task structure of the parsed spoken dialogs. The parse-based dialog model can be a shift-reduce model, a start-complete model, or a connection path model.