ML Email Classification via Rule-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning models struggle to efficiently classify and route emails and documents based on their actionability, leading to human operators spending time reviewing unactionable messages.
Innovation Solution
A system that uses a machine learning model to classify sentences within emails or documents as actionable or unactionable, and routes them accordingly, with the option to bypass the machine learning classifier using a rule-based system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional machine learning models are used to classify all emails, then classification coverage is improved, but processing time and computational resources are wasted on unactionable messages
Solution Approach 1:
The patent segments the email classification task into two distinct stages: first a rule-based filter categorizes emails into potentially actionable and unactionable groups, then a machine learning model is applied only to the potentially actionable subset. This segmentation resolves the contradiction by maintaining high classification coverage for relevant emails while avoiding the computational cost of applying complex ML models to all emails including clearly unactionable ones.
Solution Approach 2:
The rule-based classification system performs preliminary filtering before the machine learning model processes emails. By applying simple, fast rules first to eliminate obviously unactionable emails, the system prepares the data in advance so that the more computationally intensive ML model only needs to process a reduced subset, thereby reducing overall processing time while maintaining accurate classification.
2Measurement precision
If machine learning models process all incoming messages, then classification accuracy is improved, but operational efficiency deteriorates due to reviewing unactionable messages
Solution Approach 1:
The system segments message processing into two pathways: a fast rule-based filter for initial triage and a ML-based classifier for detailed analysis of only relevant messages. This segmentation maintains high classification accuracy for actionable messages while improving operational efficiency by preventing human operators from reviewing unactionable messages that would be incorrectly classified or require manual verification.
Solution Approach 2:
The rule-based classification system acts as an intermediary between the incoming message stream and the machine learning model. This intermediary layer pre-processes messages using simple, fast rules to identify and filter out unactionable messages before they reach the ML model, thereby protecting the overall system from processing waste and improving operational efficiency while the ML model maintains high accuracy on relevant messages.
3Speed
If a rule-based system is used for email classification, then processing speed is improved, but classification accuracy deteriorates
Solution Approach 1:
The classification system is segmented into two layers with different performance characteristics: a rule-based layer optimized for speed that handles the bulk of filtering, and a machine learning layer optimized for accuracy that handles the remaining potentially actionable messages. This segmentation allows the system to achieve both high processing speed (through the rule-based filter) and high classification accuracy (through the ML model on relevant messages).
Solution Approach 2:
The rule-based system performs partial classification action on all messages, quickly identifying and filtering out clearly unactionable ones. While the rule-based system alone would have insufficient accuracy, it performs enough classification work to reduce the message volume to a level where the ML model can achieve high accuracy on the remaining subset, thereby achieving both speed and accuracy at the system level.
Data Source
AI summary
Methods, systems, and apparatus, including computer programs encoded on computer storage media, for machine learning. One of the methods includes receiving a message; parsing the message to identify individual sentences within the message; encoding each sentence into a vector that represents a semantic meaning of the sentence; applying one or more classification models to the message including: providing each vector to a machine learning model, wherein the machine learning model is trained to generate a prediction of one or more classifications for each sentence of the message; generating a predicted classification for the message as a whole; and determining a message routing based on the predicted classification.