Conversational Search Interface With Confidence-Based Clarification

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing conversational chatbots face challenges in delivering high-quality and accurate responses across multiple channels, leading to inefficiency and user frustration due to the separation of general and user-specific information sources.

Innovation Solution

An integrated, multi-channel conversational utility that trains a machine-learning model to associate input keywords with both general and user-specific information sources, providing a single search icon to receive inquiries and generating results with confidence scores, and employing follow-up questions when necessary to clarify intent.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If multiple separate information sources (general and user-specific) are used, then comprehensive information coverage is improved, but system complexity and user confusion increase

Engineering Contradiction:

Improveinformation coverageVSAvoidsystem complexity

Core Design Contradiction:

Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent merges general information sources and user-specific information sources into a single unified information architecture. The system combines multiple data sources behind a single search icon, integrating general knowledge base and user-specific data without requiring users to navigate separate channels or understand complex source differentiation.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The unified search interface serves multiple functions simultaneously - it queries both general information sources and user-specific information sources through a single interaction point. The search icon acts as a universal entry point that handles diverse information retrieval needs without requiring separate specialized interfaces for each information type.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Quantity of substance

If multiple separate information sources are used, then comprehensive information coverage is improved, but user frustration and confusion increase

Engineering Contradiction:

Improveinformation coverageVSAvoiduser experience

Core Design Contradiction:

Quantity of substanceVSEase of operation

Solution Approach 1:

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The system automatically determines which information sources are relevant to each user based on their profile and query context. The machine learning model autonomously selects and retrieves appropriate information from either general or user-specific sources without requiring user input or awareness of the underlying source structure, making the system self-adaptive and user-friendly.

Inventive Principle:

Principle #25Self-service

3Ease of operation

If a single search icon is used for all inquiries, then ease of operation is improved, but response accuracy may deteriorate

Engineering Contradiction:

Improveinterface simplicityVSAvoidresponse accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating information retrieval strategies based on user-specific characteristics and query context. While the interface remains uniform, the underlying system adapts its search behavior locally for each user and query type, selecting appropriate information sources and weighting results based on individual user profiles and situational context.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system dynamically changes search parameters and retrieval strategies based on user profile data, query history, and contextual information. The machine learning model adjusts information source selection and result ranking in real-time based on observed user behavior patterns, ensuring high response accuracy while maintaining interface simplicity.

Inventive Principle:

Principle #35Parameter changes

4Measurement precision

If machine learning model generates results with confidence scores, then response accuracy is improved, but processing time increases

Engineering Contradiction:

Improveresponse accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent applies partial action by retrieving and processing only the necessary information sources based on confidence score thresholds. When confidence scores indicate high relevance, the system uses pre-computed results without extensive additional processing. The machine learning model selectively queries information sources based on predicted relevance, avoiding unnecessary processing time for low-confidence queries while ensuring accuracy for high-confidence results.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12468718B1Integrated multi-channel conversational utility

Publication Date: 2025.11.11 BANK OF AMERICA CORP

AI summary

Systems and methods for an integrated, multi-channel, conversational utility are provided. Methods include providing a single search icon on a user device accessing an online portal. Methods include receiving a conversational input inquiry and, via a specially trained ML model, generating a first set of results including at least one general information resource result and at least one user-specific information result. When the first set of results exceeds a threshold confidence score, methods include displaying the first set of results as a response to the input inquiry. When the confidence score fails to exceed the threshold score, methods include generating a conversational follow-up question designed to clarify the intent of the input inquiry, receiving a response to the follow-up question, generating a second set of results and, when the second set of results exceeds the threshold confidence score, displaying the second set of results as the input inquiry response.