Multi-Intent Utterance Segmentation via Synthetic Data Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern virtual assistants implemented with a rules-based approach are limited in addressing queries and commands that exceed the scope of predetermined rules, failing to provide meaningful responses for unstructured user inputs.
Innovation Solution
A method for synthesizing training data for multi-intent utterance segmentation in machine learning-based dialogue systems, which involves creating a multi-intent training corpus by combining in-domain and out-of-domain utterances with conjunction terms, using probabilistic selection to form synthetic multi-intent utterances, and training models to classify and segment input utterances into single-intent components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a rules-based approach is used to implement virtual assistants, then the system can provide reliable responses to specific predetermined queries, but it fails to address unstructured user inputs that exceed the scope of predefined rules
Solution Approach 1:
The patent replaces the mechanical rules-based system with a machine learning-based natural language processing system. The virtual assistant uses trained models to automatically understand and respond to unstructured user inputs without requiring predefined rules for each query type, thereby maintaining reliability while significantly improving adaptability to new and varied queries.
Solution Approach 2:
The system changes the operational parameters from fixed rules to dynamic machine learning models that can adapt to different query types. By using probabilistic models and neural networks, the system can handle a broader range of queries while maintaining response quality, resolving the contradiction between reliability and versatility.
2Adaptability or versatility
If machine learning models are trained with synthetic multi-intent utterances formed by probabilistically combining in-domain and out-of-domain corpus data, then the model's ability to handle diverse queries improves, but the training data complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-processing and organizing training data into structured corpora (in-domain and out-of-domain) before model training. By preparing synthetic multi-intent utterances in advance with proper labeling and segmentation, the system reduces the complexity burden during actual training while improving the model's ability to handle diverse queries.
Solution Approach 2:
The training data is segmented into distinct components (in-domain corpus, out-of-domain corpus, intent labels, segmentation tags) that can be independently processed and combined. This segmentation approach manages training data complexity by breaking down the complex synthesis process into manageable, reusable components.
Data Source
AI summary
Systems and methods for synthesizing training data for multi-intent utterance segmentation include identifying a first corpus of utterances comprising a plurality of distinct single-intent in-domain utterances; identifying a second corpus of utterances comprising a plurality of distinct single-intent out-of-domain utterances; identifying a third corpus comprising a plurality of distinct conjunction terms; forming a multi-intent training corpus comprising synthetic multi-intent utterances, wherein forming each distinct multi-intent utterance includes: selecting a first distinct in-domain utterance from the first corpus of utterances; probabilistically selecting one of a first out-of-domain utterance from the second corpus and a second in-domain utterance from the first corpus; probabilistically selecting or not selecting a distinct conjunction term from the third corpus; and forming a synthetic multi-intent utterance including appending the first in-domain utterance with one of the first out-of-domain utterance from the second corpus of utterances and the second in-domain utterance from the first corpus of utterances.