Keyword Extraction Using Gaussian Mixture TF-IDF
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional keyword extraction methods, such as TF-IDF based on Bag of Words, struggle to effectively identify representative words from texts classified into clusters, especially in contexts like customer service where different individuals use varying wording, and fail to capture the meaning of words.
Innovation Solution
The method extends TF-IDF to distributed representation by vectorizing word sets, training Gaussian mixture distribution models for both overall and cluster-specific data, calculating TF-IDF values based on term frequency and inverse document frequency, and rearranging word sets in descending order to extract keywords.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If Bag of Words representation is used for TF-IDF calculation, then the method is simple and computationally efficient, but it cannot capture the semantic meaning of words and fails to handle synonymous expressions
Solution Approach 1:
The patent transforms the representation parameters of words from discrete Bag of Words vectors to continuous distributed representation vectors (embeddings). This parameter change enables the system to capture semantic meanings and synonymous relationships while maintaining compatibility with TF-IDF calculation framework, thus resolving the contradiction between simplicity and information preservation
Solution Approach 2:
The patent replaces the traditional mechanical counting-based Bag of Words system with a semantic-aware distributed representation system. By substituting the underlying representation mechanism while preserving the TF-IDF calculation structure, the system gains semantic understanding capabilities without completely abandoning the original simple framework
2Productivity
If conventional TF-IDF method is applied to clustered text data, then keyword extraction can be performed, but it cannot effectively identify representative words when different persons use different wordings for the same meaning
Solution Approach 1:
The patent makes the TF-IDF framework universal by enabling it to handle both exact word matches and semantic equivalents. The distributed representation layer acts as a universal translator that maps different wordings to common semantic concepts, allowing the keyword extraction method to adapt to various expression styles while maintaining its core functionality
Solution Approach 2:
The patent introduces distributed representation vectors as an intermediary between the raw text and the TF-IDF calculation. This intermediary layer captures semantic meanings and bridges the gap between different wordings, enabling the system to identify representative words across synonymous expressions while preserving the original TF-IDF extraction capability
3Loss of information
If distributed representation is used for keyword extraction, then semantic meaning of words can be captured, but the method becomes more complex and computationally intensive
Solution Approach 1:
The patent performs preliminary vectorization of the corpus into distributed representations before applying TF-IDF calculation. By pre-processing the text data into semantic vectors upfront, the system avoids the need for complex real-time semantic analysis during keyword extraction, thus reducing overall computational complexity while preserving semantic information
Solution Approach 2:
The patent segments the keyword extraction process into distinct stages: (1) text vectorization using distributed representation, (2) TF-IDF calculation on vectorized data, and (3) keyword selection. This segmentation allows each stage to be optimized independently, managing complexity by breaking down the overall process into manageable components
Data Source
AI summary
A computer-implemented method, a computer program product, and a computer system for keyword extraction. A computer vectorizes word sets by splitting text in a corpus. The computer trains a Gaussian mixture distribution model for the word sets and obtains an inverse document frequency (IDF) value of a word set from the learned Gaussian mixture distribution model. The computer trains a Gaussian mixture distribution model for a cluster and obtains a term frequency (TF) value of the word set in the cluster from the learned Gaussian mixture distribution model. The computer calculates a term frequency-inverse document frequency (TF-IDF) value of the word set in the cluster, based on the TF value and the IDF value. The computer calculates TF-IDF values of the word sets in clusters and rearranges the word sets in a descending order of the TF-IDF values to obtain extracted keywords.