Voice Interface Fuzzy Matching for Imprecise Entity Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional voicebots struggle with conversational and real-world caller dialog due to misinterpretation, confusion from pauses, imprecise identifications, and side conversations, and lack effective performance evaluation methods.
Innovation Solution
An intelligent voice interface system that includes audio handling to filter irrelevant audio, handles out-of-sequence dialog, infers user states, and translates voice communications, combined with a call review tool for manual review and improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional voicebots use strict menu-driven IVR systems, then the system structure is simple and easy to control, but the system cannot handle conversational and real-world caller dialog effectively
Solution Approach 1:
The system segments the caller's speech into multiple segments and compares each segment against multiple possible responses independently. This allows the voicebot to handle conversational dialog by processing discrete units of information separately, improving adaptability while maintaining manageable complexity through modular processing.
Solution Approach 2:
The system dynamically adjusts its response based on the caller's actual speech patterns, pauses, and conversational flow rather than forcing a rigid menu structure. The voicebot adapts its behavior in real-time based on detected user state, enabling it to handle conversational dialog while managing complexity through adaptive rather than static processing.
2Reliability
If conventional voicebots require highly ordered sequence of caller inputs, then the evaluation process is simple, but the system becomes confused by real-world calling behaviors like pauses and stalling language
Solution Approach 1:
The system performs preliminary actions by detecting and filtering out irrelevant audio segments (such as pauses, stalling language, and non-informative utterances) before processing the actual caller input. This preliminary cleaning of the audio stream ensures that only relevant information is analyzed, improving reliability in understanding caller intent while maintaining ease of operation by automatically handling real-world calling behaviors.
Solution Approach 2:
The system extracts and removes irrelevant portions of the audio signal (pauses, stalling language, background noise) from the processing stream. By taking out these distracting elements, the voicebot can focus on extracting only the meaningful caller input, improving accuracy in understanding while preserving the flexibility needed for natural conversation.
3Measurement precision
If conventional voicebots use basic string matching, then the implementation is simple, but the system fails to accurately match imprecise caller identifications like 'a '04 Chevy' rather than '2004 Chevrolet Silverado 1500'
Solution Approach 1:
The system segments the caller's identification into multiple segments (e.g., year, make, model) and compares each segment against the database independently. This segmentation allows fuzzy matching to work effectively on partial information like 'a '04 Chevy' by matching individual segments rather than requiring exact full-string matches, improving measurement precision while managing complexity through modular comparison.
Solution Approach 2:
The system performs partial matching by comparing only the relevant portions of the caller's identification against the database records. Instead of requiring complete exact matches, the system uses partial string matching on individual segments, allowing it to successfully match imprecise identifications while maintaining manageable algorithm complexity through selective comparison.
4Loss of information
If manual review of problem calls is performed, then the evaluation process provides detailed insights, but the evaluation process becomes very time consuming and subjective
Solution Approach 1:
The system performs self-service by automatically analyzing its own performance using the same segmentation and fuzzy matching algorithms that process caller input. The system automatically identifies problem calls and generates evaluation data without requiring manual reviewer intervention, thereby providing detailed performance insights while eliminating the time-consuming and subjective nature of manual evaluation.
Solution Approach 2:
The system implements feedback by automatically evaluating its own performance based on the same processing algorithms used for caller interactions. The system generates self-diagnostic data about where it fails to accurately match entities or understand caller input, providing objective performance insights while eliminating manual evaluation time through automated self-assessment.
Data Source
AI summary
A method for identifying entities may include, during a voice communication with a caller via a caller device, sending to the caller device a first voice prompt that asks the caller to identify a particular entity, receiving from the caller device caller input data indicative of a voice response of the caller, and analyzing the caller input data to determine a set of words spoken by the caller. The method may also include, for each segment of two or more segments of the set of words, determining a level of string matching between the segment and a corresponding segment in a record stored in a database, determining, and based upon the levels of string matching, a level of match certainty for the particular entity from among at least three possible levels of match certainty, and/or selecting, based upon the level of match certainty, a pathway of the algorithmic dialog.