Category-Topic Model for Semantic Text Summarization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text summarization services often rely on non-human-readable topics that lack semantic meaning, making it difficult to accurately generate summaries that reflect the semantic insights of textual data.
Innovation Solution
The development of a category-topic model that determines co-occurrence relationships between topics and category labels, and between sequences of words and topics, allowing for the generation of human-readable summaries by sampling and calculating probabilities to rank relevant sequences of words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If non-human-readable topics are used for text summarization, then the summarization process can be automated, but the semantic meaning and insight into the textual data are lost
Solution Approach 1:
The patent introduces category labels as intermediary elements that bridge the gap between automated processing and semantic understanding. These category labels serve as human-readable representations that capture the semantic meaning of text segments while enabling automated summarization processes to function effectively.
Solution Approach 2:
The patent changes the representation parameters of topics from non-human-readable abstract vectors to human-readable category labels. This parameter transformation allows the system to maintain automation while providing semantic interpretability, as the category labels directly reflect the semantic categories of the text content.
2Loss of information
If category labels are introduced to improve semantic understanding, then the semantic meaning is preserved, but the complexity of the system increases
Solution Approach 1:
The patent segments the text into discrete segments and assigns category labels to each segment independently. This segmentation approach allows the system to handle semantic meaning at a granular level without requiring complex global processing, thereby reducing overall system complexity while preserving semantic information.
Solution Approach 2:
The category labels serve multiple functions simultaneously: they provide semantic interpretation, enable automated ranking, facilitate summarization, and maintain consistency across different text documents. This multi-functionality reduces the need for separate specialized components, thereby reducing system complexity.
3Measurement precision
If sampling and probability calculations are performed to rank sequences, then the accuracy and precision of summarization improve, but the processing time increases
Solution Approach 1:
The patent performs sampling and probability calculations in advance during the training phase to pre-compute ranking scores for different text sequences. This preliminary action allows the system to make accurate summarization decisions during actual use without performing time-consuming calculations in real-time, thus improving accuracy while reducing processing time.
Solution Approach 2:
The patent implements continuous sampling and probability estimation processes that run concurrently with other summarization tasks. This continuous execution of useful actions ensures that accurate ranking is maintained while minimizing idle processing time, as the sampling processes utilize available computational resources efficiently throughout the operation.
Data Source
AI summary
Provided are methods, systems and apparatus which include computer program products, for generating topic models for text summarization In one aspect, a method includes receiving a first document of text that is associated with one or more category labels and that includes one or more sequences of one or more words, determining a category label that represents a first category associated with the first document, sampling the one or more sequences to determine a topic and a co-occurrence relationship between the topic and the category label, where a topic represents a subdivision within a category, sampling the one or more sequences to determine a co-occurrence relationship between a sequence in the first document and the topic, and generating a category-topic model that represents the co-occurrence relationships.