RNN Abstractive Summarization for Unstructured Interaction Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abstractive summarization techniques are inadequate for handling disorganized and entropic content characteristic of interaction data generated by networks of individuals interacting over time, as they are not trained or tested on such data.
Innovation Solution
An improved computational method using a recurrent neural network (RNN) implemented on an Object Translator computer system to encode interaction data into an intermediate representation and select the corpus with the lowest cost of generating a query, effectively summarizing and characterizing entities from unstructured text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional extractive summarization methods are used on interaction data, then processing speed is maintained, but accuracy of summarization deteriorates due to inability to handle disorganized and entropic content
Solution Approach 1:
The patent replaces traditional mechanical extractive summarization methods with a neural network-based abstractive summarization system. The neural network learns to generate summaries by understanding the semantic meaning of interaction data, rather than relying on fixed rules for selecting important sentences. This substitution enables accurate handling of disorganized interaction data while maintaining system feasibility through standardized neural network architectures.
2Adaptability or versatility
If abstractive summarization is implemented without proper training on interaction data, then summarization quality improves, but applicability to interaction data deteriorates
Solution Approach 1:
The patent applies preliminary training actions to the neural network using interaction data before deployment. The system is pre-trained on interaction data to learn the specific patterns and characteristics of such data, enabling it to generalize well when applied to new interaction datasets. This preliminary action reduces the need for extensive customization and improves adaptability.
Solution Approach 2:
The neural network is designed with universal architecture that can handle various types of interaction data through a single unified model. The system uses standardized input processing and attention mechanisms that work across different interaction scenarios, reducing the need for separate specialized models for different data types.
3Loss of information
If more sentences are selected for summarization, then coverage of content improves, but loss of time in processing increases
Solution Approach 1:
The neural network processes interaction data continuously, generating summaries in a single pass through the data. The attention mechanism allows the model to focus on relevant portions of the input data dynamically, maintaining continuous processing without needing to repeatedly analyze the same content. This continuous action ensures complete content coverage while minimizing processing time.
Solution Approach 2:
The system uses attention mechanisms to selectively focus on the most relevant portions of interaction data, processing only the necessary information in detail while summarizing other parts more efficiently. This partial action approach ensures that critical content is thoroughly covered while reducing overall processing time by avoiding exhaustive analysis of all data points.
Data Source
AI summary
A recurrent neural network (RNN) method implemented on a computer system is used to produce summaries of unstructured text generated by multiple networks of individuals interacting over time by encoding the unstructured text into intermediate representations and decoding the intermediate representations into summaries of each network. Parameter data for the RNN is obtained by using multiple different versions of the same source texts to train the computer system. The method and computer system can be used to identify which of the networks match a query by determining which network generates the query with low or lowest cost.