Phoneme String Information for Conversation Situation Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Call centers face challenges in detecting specific conversation situations between operators and customers due to the difficulty in setting exhaustive keywords and varying expressions of similar meanings, which affects the accuracy and efficiency of conversation analysis.
Innovation Solution
An information processing apparatus that generates phoneme string information from voice data and uses machine learning to learn network parameters, enabling the detection of predetermined conversation situations without requiring exhaustive keyword setting, by employing LSTM and DNN models to differentiate between normal and abnormal conversation vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If keywords are exhaustively set to detect specific conversation situations, then detection coverage is improved, but device complexity and difficulty of operation increase due to the need to set numerous keywords manually
Solution Approach 1:
The system automatically learns and extracts keywords from conversation data through machine learning models (LSTM and DNN), eliminating the need for manual keyword setting. The model self-adapts to identify relevant conversation situations by training on labeled data, where keywords are discovered autonomously rather than being predetermined by operators.
Solution Approach 2:
The system transforms the approach from static keyword matching to dynamic parameter-based detection. By using phoneme string information as input parameters and training the model to learn optimal detection parameters automatically, the system adapts its detection criteria based on the learned patterns in conversation data rather than relying on fixed keyword lists.
2Reliability
If manual keyword setting is performed to detect conversation situations, then detection capability is improved, but loss of time increases due to trial and error in setting keywords
Solution Approach 1:
The system performs preliminary learning by training the machine learning model on labeled conversation data before actual detection begins. This preliminary training phase automatically establishes the detection capabilities, eliminating the need for time-consuming trial and error keyword setting that would otherwise be required during operation.
Solution Approach 2:
The model automatically learns optimal detection parameters and patterns from training data, performing the keyword equivalent work autonomously without human intervention. This self-learning process replaces the manual trial-and-error keyword setting process, significantly reducing the time investment required.
3Ease of operation
If phoneme string information is used instead of keywords, then ease of operation is improved, but measurement precision may worsen due to loss of semantic information
Solution Approach 1:
The system introduces phoneme string information as an intermediary representation between raw voice data and semantic meaning. Rather than directly processing keywords or full text, the model uses phoneme sequences as a intermediate layer that preserves sufficient information for detection while simplifying the input processing and enabling automatic learning without manual keyword specification.
Data Source
AI summary
An information processing apparatus includes a memory; and a processor coupled to the memory and the processor configured to: generate phoneme string information in which a plurality of phonemes included in voice information is arranged in time series, based on a recognition result of the phonemes for the voice information; and learn parameters of a network such that when the phoneme string information is input to the network, output information that is output from the network approaches correct answer information that indicates whether a predetermined conversation situation is included in the voice information that corresponds to the phoneme string information.