Satisfaction State Estimation Model Using Transition Weights

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies for estimating customer satisfaction in call centers do not account for changes in satisfaction states over time, failing to consider the time series correlation of customer satisfaction during conversations.

Innovation Solution

A state-of-satisfaction change pattern model and estimation model are developed using transition weights in a state sequence to estimate the posterior probability of utterance features, allowing for the consideration of changes in customer satisfaction patterns.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If existing technologies estimate satisfaction from features at given time points only, then the estimation process is simple, but the accuracy of satisfaction assessment is insufficient because time series correlation is ignored

Engineering Contradiction:

Improveaccuracy of satisfaction assessmentVSAvoidcomplexity of estimation model

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies dynamics by transitioning from static satisfaction estimation at discrete time points to dynamic satisfaction state tracking that models temporal evolution. The hidden Markov model captures the dynamic nature of customer satisfaction by representing it as a sequence of states that evolve over time, allowing the system to adapt to changing satisfaction levels throughout a conversation.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent adds a temporal dimension to satisfaction estimation by introducing time series analysis. Instead of estimating satisfaction independently at each time point, the system incorporates the temporal dimension through transition probabilities between satisfaction states, enabling the model to consider how satisfaction evolves over time rather than treating each moment in isolation.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Measurement precision

If time series correlation of satisfaction states is considered, then the accuracy of satisfaction estimation is improved, but the complexity of the model increases

Engineering Contradiction:

Improveaccuracy of satisfaction estimationVSAvoidcomplexity of state transition modeling

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent introduces transition probabilities as an intermediary mechanism that bridges the gap between simple point-in-time estimation and complex temporal modeling. These transition probabilities serve as pre-computed intermediaries that encapsulate the temporal relationships between satisfaction states, allowing the system to leverage time series correlation without directly implementing complex temporal dependencies during real-time estimation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent applies preliminary action by pre-computing transition probabilities between satisfaction states during an offline training phase. This preliminary computation of temporal relationships allows the online estimation system to simply apply these pre-established probabilities rather than performing complex temporal analysis in real-time, thus improving accuracy without proportionally increasing online computational complexity.

Inventive Principle:

Principle #10Preliminary action

3Adaptability or versatility

If state transition sequences are modeled for satisfaction changes, then the ability to capture satisfaction dynamics is improved, but the difficulty of detecting and measuring satisfaction states increases

Engineering Contradiction:

Improveability to capture satisfaction dynamicsVSAvoiddifficulty of detecting satisfaction state changes

Core Design Contradiction:

Adaptability or versatilityVSDifficulty of detecting and measuring

Solution Approach 1:

The patent implements feedback by using the output satisfaction state at one time point as input for determining the transition probability to the next state. This feedback mechanism allows the system to continuously update its understanding of satisfaction dynamics based on previous states, improving adaptability to different conversation patterns while maintaining a systematic approach to state detection through the structured HMM framework.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11521641B2Model learning device, estimating device, methods therefor, and program

Publication Date: 2022.12.06 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

State-of-satisfaction change pattern models each including a set of transition weights in state sequences of the states of satisfaction are obtained for predetermined change patterns of the states of satisfaction, and a state-of-satisfaction estimation model for obtaining the posteriori probability of the utterance feature amount given the state of satisfaction of an utterer is obtained by using the utterance-for-learning feature amount and a correct value of the state of satisfaction of an utterer who gave an utterance for learning corresponding to the utterance-for-learning feature amount. By using the input utterance feature amount and the state-of-satisfaction change pattern models and the state-of-satisfaction estimation model, an estimated value of the state of satisfaction of an utterer who gave an utterance corresponding to the input utterance feature amount is obtained.