Sliding-Window N-Gram Protocol for IVR Entity Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Interactive Voice Response (IVR) systems face challenges in accurately extracting verifiable entities from human utterances, which hinders their ability to understand user intent and provide effective responses, leading to suboptimal customer experience.
Innovation Solution
The method involves canonicalizing user-utterances into tokens, generating n-gram sequences using a sliding-window protocol, processing them in parallel threads to verify entities, and utilizing a database of entity-verifiers to determine valid entities, ensuring efficient extraction without significantly increasing computation time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional entity extraction methods are used in IVR systems, then the system can process user utterances, but the entity extraction accuracy is insufficient and cannot reliably understand user intent
Solution Approach 1:
The patent segments the user utterance into multiple n-gram sequences (unigrams, bigrams, trigrams, etc.) of different lengths. Each sequence is processed independently through parallel threads to extract entities, allowing the system to capture entities at various granularities and improve overall extraction accuracy and reliability.
Solution Approach 2:
The patent generates a comprehensive number of n-gram sequences that exceeds the minimum needed for entity extraction. By creating multiple overlapping sequences of different lengths and processing all of them in parallel, the system performs excessive analysis to ensure no potential entity is missed, thereby improving extraction reliability.
2Measurement precision
If comprehensive n-gram sequences are generated and processed in parallel threads, then entity extraction accuracy improves to over 99%, but computation time may increase
Solution Approach 1:
The patent transitions from sequential processing to parallel processing by introducing a new dimension of concurrency. Multiple threads execute simultaneously, each handling different n-gram sequences, thereby maintaining high accuracy while reducing the time penalty that would otherwise result from comprehensive analysis.
Solution Approach 2:
The patent creates multiple copies of the verification process running in parallel threads. Instead of one sequential verification pass, the system spawns identical verification logic across multiple threads, each processing different n-gram sequences simultaneously, achieving both high accuracy and efficient time utilization.
3Reliability
If multiple n-gram sequences are processed contiguously in parallel threads, then entity verification reliability improves, but system complexity increases
Solution Approach 1:
The patent implements a universal verification mechanism that handles all n-gram sequences through a common entity verifier component. This multi-functional approach maintains high reliability by consistently applying the same verification logic across different sequence types while avoiding the need for separate complex processing paths for each n-gram length.
Solution Approach 2:
The patent introduces an entity verifier as an intermediary component between n-gram sequence generation and final entity extraction. This mediator standardizes the verification process, managing the complexity of parallel thread coordination and entity validation in a unified manner, thereby improving reliability without proportionally increasing overall system complexity.
Data Source
AI summary
A system for extracting verifiable entities from a user-utterance received on an automated calling service is provided. The system may include a receiver configured to receive a user-utterance, a processor and a non-transitory computer-readable media comprising computer-executable instructions. The processor may be configured to execute the instructions which, canonicalize the user-utterance into a plurality of tokens, determine the number of tokens of the user-utterance, and generate, using a sliding-window protocol, a comprehensive number of n-gram sequences from the user-utterance. The processor may be configured to process a plurality of threads of execution that may include a series of actions executed on the n-gram sequences to identify and extract verified entities from the user-utterance.