Intent Vector Matching for Conversational Agent Updates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated conversational agents require significant manual effort to update intents, including adding new intents, splitting existing ones, or deleting them, which is inefficient and time-consuming, especially when analyzing large conversation datasets or intent pools.
Innovation Solution
A method and apparatus that generate vector representations of conversational inputs using an embedding model to determine intent data, compare these representations with a data pool to retrieve similar sentences, and update intent data efficiently, reducing the need to analyze entire datasets or intent pools.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis of huge conversation datasets is performed to add new intents, then intent accuracy is improved, but time consumption and manual effort increase significantly
Solution Approach 1:
The patent applies partial action by retrieving only a subset of sentences from the data pool that are most relevant to the new intent, rather than analyzing the entire dataset. The system retrieves sentences containing keywords or semantically similar to the new intent definition, processes only this partial set for annotation, and uses them to create the intent. This dramatically reduces the time and effort required while maintaining high intent accuracy.
2Measurement precision
If entire intent pools are manually reviewed to split or delete intents, then intent precision is improved, but productivity decreases due to extensive time required
Solution Approach 1:
The patent extracts only the necessary portion of the intent pool for analysis. When splitting or deleting an intent, the system retrieves sentences that are specifically related to the intent in question using keyword matching and semantic similarity, rather than requiring manual review of the entire intent pool. This extraction approach maintains intent precision while significantly improving productivity.
3Reliability
If comprehensive dataset analysis is performed to ensure accurate intent classification, then response accuracy is improved, but device complexity and processing requirements increase
Solution Approach 1:
The patent segments the intent update process into distinct stages: (1) retrieve relevant sentences from the data pool based on keywords and semantic similarity, (2) annotate and classify only these retrieved sentences, (3) create or update the intent based on the annotated subset. This segmentation allows the system to achieve high response accuracy by focusing computational resources on the most relevant data, rather than processing the entire dataset, thereby reducing overall processing complexity.
Data Source
AI summary
Embodiments provide methods and apparatus for improving responses of automated conversational agents. The method includes generating a vector representation of a conversational input provided by a user. The vector representation is used to determine an intent of the conversational input. Further, annotators generate bait sentences that cover multiple aspects of the intent. Then, sentences in a data pool are accessed. The bait sentences and the data pool sentences are converted into a first and a second set of vector representations, respectively. The first and the second set of vector representations are compared to retrieve a list of similar sentences. The list of similar sentences includes one or more sentences of the data pool that are semantically similar to the bait sentences. The list of similar sentences is analyzed for updating the intent data and thereby improve the responses.