Message Distribution Destination Determination Using Co-occurrence Probability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing message distribution systems face challenges in identifying the correct distribution destination when a message contains unknown words not registered in the co-occurrence probability information, leading to incorrect classification and potential misrouting of messages.
Innovation Solution
The system extracts context words from messages containing unknown words and calculates co-occurrence probabilities based on similar usage words from messages without the unknown words, registering these probabilities to enhance the co-occurrence information for improved distribution destination determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Bayesian classification is used to determine distribution destinations based on co-occurrence probability information, then message routing accuracy is improved for known words, but the system fails to correctly classify messages containing unknown words not registered in the co-occurrence probability information
Solution Approach 1:
The patent introduces a mediator word selection unit that selects mediator words from the vocabulary set to bridge the gap between unknown words and the co-occurrence probability information. These mediator words serve as intermediaries to calculate pseudo-co-occurrence probabilities for unknown words, enabling the system to handle messages with unfamiliar terminology while maintaining routing accuracy.
Solution Approach 2:
The patent changes the parameter representation by calculating pseudo-co-occurrence probabilities instead of relying on pre-registered co-occurrence data. This parameter transformation allows the system to adapt to unknown words by computing new probability values based on mediator words, thereby extending the system's adaptability without sacrificing measurement precision for known words.
2Measurement precision
If extensive training data is collected to register co-occurrence probabilities for all possible words, then classification accuracy for known words is improved, but the system complexity and data requirements increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-registering a comprehensive vocabulary set and co-occurrence probability information for known words during system initialization. This preliminary preparation enables the system to quickly process messages with known words using Bayesian classification, while the mediator word mechanism provides a fallback for unknown words, reducing the need to pre-register every possible word and thereby lowering system complexity.
Solution Approach 2:
The patent uses copying by creating pseudo-co-occurrence probability information through mediator words when exact co-occurrence data for unknown words is unavailable. This copying approach allows the system to replicate the probabilistic reasoning process for unknown words based on related known words, maintaining classification accuracy without requiring extensive training data for every possible term.
3Adaptability or versatility
If the co-occurrence probability information is updated frequently to adapt to new words, then adaptability to new messages is improved, but the processing time and computational resources increase
Solution Approach 1:
The patent applies partial action by updating co-occurrence probability information selectively rather than comprehensively. The system updates probabilities only for mediator words and their associated unknown words when new patterns are detected, rather than reprocessing the entire vocabulary set. This selective update approach maintains adaptability to new words while significantly reducing the time and computational resources required compared to full system retraining.
Data Source
AI summary
With reference to information storing a co-occurrence probability of each of plural words in association with each of distribution-destinations, the apparatus extracts, from a message to be distributed, an unknown-word that is not included in the plural words, where the co-occurrence probability indicates a probability that each word is included in a message to be distributed to each distribution-destination. When the unknown-word is present in the message at a predetermined ratio or greater, the apparatus extracts first-words included in all of a first-group of messages each including the unknown-word, calculates the co-occurrence probability of the unknown-word for each distribution-destination, based on the information regarding the distribution-destinations to which a message including the first-words and a second-word, which is included in all of a second-group of messages each including the first-words and not including the unknown-word, is to be distributed, and registers the calculated co-occurrence probability in the information.