Intention Classification via Fixed-Length Value Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional predictive mechanisms for customer intentions are limited by their static nature, requiring fixed-length inputs and struggling to handle variable interaction attributes, such as web pages visited, which makes it difficult to accurately predict customer intentions across multiple interaction channels.
Innovation Solution
A method and apparatus that classify customer interaction data into fixed-length sequences of values, allowing the use of standard classifiers to predict customer intentions by defining categories for both static and dynamic attributes, enabling the processing of variable-length interaction data and facilitating the use of advanced classifiers like SVM and predictive association rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional predictive mechanisms use fixed-length inputs, then the system structure is simple, but the system cannot accurately handle variable interaction attributes like web pages visited
Solution Approach 1:
The patent segments the customer interaction journey into discrete action categories (e.g., web page visits, IVR options selected, chat messages sent). Each action is represented as a separate entity with its own sequence of values, allowing the system to handle variable-length interactions by processing them as discrete segments rather than requiring fixed-length inputs.
Solution Approach 2:
The patent introduces a new dimension by transforming variable-length interaction sequences into fixed-length value sequences through categorization. By mapping each action to a standardized set of categories and generating sequences of fixed length (equal to the number of action categories), the system maintains adaptability to variable interactions while providing fixed-length inputs for classification algorithms.
2Measurement precision
If the system uses advanced intention classifiers, then the prediction accuracy improves, but the system complexity increases
Solution Approach 1:
The patent introduces an intermediary layer between the variable-length interaction data and the classification algorithm. This intermediary consists of the sequence generation mechanism that transforms raw interaction data into standardized fixed-length value sequences. The intermediary handles the complexity of data transformation, allowing advanced classifiers to focus solely on prediction without dealing with variable-length input complexities.
3Adaptability or versatility
If the system processes multiple interaction channels, then the coverage of customer interactions improves, but the data processing complexity increases
Solution Approach 1:
The patent creates a universal action category framework that can accommodate multiple interaction channels (web, IVR, chat, mobile app) through a single consistent structure. By defining action categories that apply across all channels and using the same sequence generation process for each channel, the system achieves multi-channel coverage without proportionally increasing processing complexity. Each channel's interactions are processed through the same standardized pipeline.
Data Source
AI summary
A computer-implemented method and apparatus for predicting customer intentions defines a plurality of categories for classifying customer interaction data. The plurality of categories includes at least one action category for classifying information related to customer actions on interaction channels. Data signals corresponding to a customer interaction on one or more interaction channels is received. The data signals include information related to at least one customer action. A sequence of values is generated for each customer action for classifying information related to the each customer action. A value is generated corresponding to each action category to configure the sequence of values. The sequence of values is associated with a fixed length equal to a number of action categories in the at least one action category. The fixed length of the sequence of values facilitates use of one or more intention classifiers to predict an intention of the customer.