Email Classification via User Profile Mediation and ML Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic message classification systems fail to accurately associate emails with specific projects due to reliance on subject line matching, leading to missed associations between emails discussing the same project with different subject fields.
Innovation Solution
A programmatic method using machine learning techniques that identifies projects by analyzing sender and recipient users, determining common projects, and using feature sets to score candidate projects based on contextual and historical information, thereby improving classification accuracy over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subject line matching is used for classification, then the system is simple and fast, but classification accuracy deteriorates because emails with different subject lines discussing the same project are not associated
Solution Approach 1:
The patent introduces user profiles as an intermediary data structure that mediates between email messages and projects. Each user profile contains metadata about the user's role, department, and historical project associations, which serves as a bridge to connect emails to projects through user relationships rather than direct subject line matching.
Solution Approach 2:
The system performs preliminary action by pre-building user profiles with metadata before email classification occurs. Historical data about user-project relationships is extracted and stored in advance, so that when new emails arrive, the system can quickly query pre-computed user profile information rather than performing complex real-time analysis.
2Productivity
If manual classification is used, then accuracy can be maintained, but productivity deteriorates due to time-consuming manual processing
Solution Approach 1:
The system implements self-service by enabling automatic classification through machine learning models that continuously learn from user feedback. The automated system handles routine classification tasks independently, and only involves human users when the model is uncertain or when feedback is needed for retraining, thereby maintaining high throughput while preserving accuracy.
Solution Approach 2:
The patent incorporates feedback mechanisms where users can correct misclassified emails, and this feedback is used to retrain the classification model. This continuous feedback loop allows the system to maintain high accuracy by learning from its mistakes while operating automatically at scale, resolving the trade-off between automated throughput and manual accuracy.
3Loss of information
If subject line matching is used, then ease of operation is maintained, but loss of information increases because contextual relationships between emails are missed
Solution Approach 1:
The patent adds another dimension to classification by incorporating user profile metadata and historical relationship data beyond the traditional subject line field. Instead of relying solely on the email subject dimension, the system analyzes multiple dimensions including sender-recipient relationships, user roles, department affiliations, and historical project associations to comprehensively capture project context.
Data Source
AI summary
Techniques are described herein for classifying an electronic message with a particular project from among a plurality of projects. In some embodiments, first and second users associated with the electronic message are identified, and one or more first projects associated with the first user and one and more second projects associated with the second user are determined. Projects that are in common between the first projects and the second projects are determined. When only a single project is in common, the electronic message is associated with the single project. When more than a single project is in common, features associated with each of the projects found to be in common are analyzed by a machine learning model to determine the most likely project to associate with the electronic message from among the projects found to be in common.