Topic Mining via Social Action Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current analytics applications fail to accurately identify topics within conversations, treating them as monolithic texts without regard to segments, leading to inefficiencies in computational resources and reduced accuracy in topic extraction.
Innovation Solution
A computing device uses a deep learning model, specifically a sequential recurrent neural network, to classify utterances into social actions, identify transition boundaries, and group adjacent segments based on similarity, reducing computational complexity and improving accuracy by storing annotated topic segments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conversations are treated as monolithic texts without segmentation, then the analysis process is simpler, but topic identification accuracy is reduced
Solution Approach 1:
The patent divides conversations into segments based on transition boundaries detected through social action classification. Each segment represents a distinct topic or interaction phase, allowing for more accurate topic identification while maintaining manageable analysis complexity through structured decomposition of the conversation flow.
2Loss of information
If all utterances are processed in detail, then topic extraction completeness is improved, but computational resource consumption increases
Solution Approach 1:
The patent extracts and removes filler utterances and non-substantive content from the conversation sequence before topic extraction. This filtering process eliminates computational waste on irrelevant text while preserving all meaningful topic-related information, achieving efficient resource utilization without sacrificing topic extraction completeness.
3Measurement precision
If deep learning models are used for utterance classification, then classification accuracy is improved, but device complexity increases
Solution Approach 1:
The patent employs a pre-trained deep learning model for social action classification that serves as an intermediary layer between raw utterances and topic extraction. This pre-trained model has already learned nuanced linguistic patterns during training, enabling accurate classification with reduced computational overhead during actual conversation analysis, thus balancing accuracy with operational efficiency.
4Measurement precision
If transition boundaries are identified to create segments, then topic segmentation accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs social action classification and transition boundary identification as preliminary steps before actual topic extraction. By pre-processing the conversation to establish segment boundaries and classify utterances in advance, the system enables faster topic extraction in subsequent operations, trading initial processing time for improved overall efficiency and accuracy.
Data Source
AI summary
A method and system of automatically identifying topics of a conversation are provided. An electronic data package comprising a sequence of utterances between conversation entities is received by a computing device. Each utterance is classified to a corresponding social action. One or more utterances in the sequence are grouped into a segment based on a deep learning model. A similarity of topics between adjacent segments is determined. Upon determining that the similarity is above a predetermined threshold, the adjacent segments are grouped together. A transcript of the conversation including the grouping of the adjacent segments is stored in a memory.