Expert Matching via ML Filtering for Communication Sessions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in accurately and reliably connecting client devices for communication sessions, especially as the number of users and device specifications increase, due to issues with granular user specifications and jargon or slang in queries.
Innovation Solution
A data processing system uses a machine learning model trained on expert records to identify relevant expert identifiers by processing queries through natural language processing, ensuring that only available and qualified experts are connected for communication sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the number of users and device specifications increase, then the system capacity increases, but the connection accuracy and reliability decrease due to granular user specifications and jargon in queries
Solution Approach 1:
The patent introduces a data processing system as an intermediary between users and client devices. This system includes a machine learning model that processes queries, filters expert identifiers, and transmits refined queries to client devices. The intermediary handles the complexity of granular specifications and jargon by translating them into actionable connection parameters, thereby maintaining high connection accuracy despite increased system capacity.
Solution Approach 2:
The patent implements preliminary action by pre-processing queries through a machine learning model before transmitting them to client devices. The system determines confidence scores for expert identifiers in advance and filters them based on threshold criteria. This preliminary filtering ensures that only relevant and reliable expert identifiers are transmitted, maintaining connection accuracy even as the number of users and specifications increases.
2Adaptability or versatility
If more expert identifiers are transmitted to client devices, then the coverage of expert matching increases, but the computational resources and network bandwidth consumption increase
Solution Approach 1:
The patent extracts only the necessary expert identifiers from the full set of available experts. The machine learning model processes queries and returns a filtered subset of expert identifiers based on confidence scores and relevance thresholds. This extraction ensures that client devices receive only the most relevant expert identifiers, reducing computational resources and network bandwidth consumption while maintaining adequate expert matching coverage.
Solution Approach 2:
The patent changes the parameters of expert identifier transmission by using confidence scores as a filtering criterion. The system adjusts the threshold confidence score to control the number of expert identifiers transmitted. This parameter change allows the system to optimize between expert matching coverage and computational resource consumption dynamically based on query complexity and system state.
3Measurement precision
If a machine learning model is used to determine confidence scores, then the prediction accuracy increases, but the system complexity increases
Solution Approach 1:
The patent positions the machine learning model as an intermediary component within the data processing system. Rather than integrating complex ML algorithms directly into client devices, the system uses a centralized ML model that processes queries and returns simplified confidence scores. This intermediary approach achieves high prediction accuracy while keeping individual device complexity manageable.
4Quantity of substance
If client devices query all expert identifiers, then the completeness of expert search increases, but the time required for connection increases
Solution Approach 1:
The patent performs preliminary action by pre-ranking and filtering expert identifiers before they are transmitted to client devices. The machine learning model processes queries in advance and returns a sorted list of expert identifiers based on confidence scores. This preliminary organization allows client devices to connect more quickly without sacrificing search completeness, as the most relevant experts are already prioritized.
Solution Approach 2:
The patent segments the expert identifier list into prioritized groups based on confidence scores. Instead of transmitting all expert identifiers equally, the system segments them into high-confidence and low-confidence groups. Client devices can focus initial connection efforts on high-confidence experts, reducing connection time while maintaining the ability to search complete expert lists if necessary.
Data Source
AI summary
Connecting devices for communication sessions is provided. A system can receive, from a client device, a first query comprising text. The system can determine, responsive to the first query and via a machine learning model, confidence scores corresponding to the expert identifiers. The system selects a first subset of expert identifiers based on a threshold confidence score or a ranking technique. The system transmits a second query based on the first query to client devices corresponding to the first subset of expert identifiers. The system identifies, based on responses to the second query, a second subset of expert identifiers. The system provides, to the client device, an indication of the second subset of expert identifiers to cause the client device to establish a communication session with at least one client device of at least one of the second subset of expert identifiers.