Attention CNN Pseudo-Code Generation from Non-English Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods face challenges in accurately and efficiently generating pseudo-code from non-English text documents, particularly in translating and summarizing business processes, workflows, and log reports into English while optimizing network resources and ensuring high accuracy for reliable process flow mapping.
Innovation Solution
A computing platform employing a neural machine translation model, attention-based convolutional neural networks (CNNs), and natural language generation models to translate, summarize, and generate flowcharts and pseudo-code from non-English text documents, utilizing inter-attention and intra-attention nets for contextual understanding and abstractive summarization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional translation and summarization methods are used for non-English documents, then processing speed is slow, but using advanced neural network models improves speed and accuracy while consuming more computational resources
Solution Approach 1:
The system segments the document processing into distinct stages: translation from non-English to English, abstractive summarization to extract key steps, flowchart generation to visualize process flow, and pseudo-code generation to create executable code. Each stage uses specialized models optimized for that specific task, improving overall efficiency while managing computational resources effectively.
Solution Approach 2:
The patent introduces intermediate representations (English translation, abstractive summary, flowchart) as mediators between the original non-English document and the final pseudo-code output. These intermediaries break down the complex translation-to-code task into manageable steps, each handled by appropriate neural network models, balancing accuracy and resource usage.
2Reliability
If accurate translation and summarization are prioritized for reliable process flow mapping, then processing time increases, but faster methods may reduce accuracy
Solution Approach 1:
The patent replaces traditional mechanical translation and summarization systems with neural machine translation models and attention-based CNNs. These neural systems learn contextual relationships and semantic meanings, providing more accurate translations and summaries that reliably capture process flow steps without excessive processing time.
Solution Approach 2:
The system adjusts key parameters including attention weights in the CNN to focus on important words and phrases, vocabulary distribution to handle domain-specific terminology, and context vector dimensions to capture sufficient semantic information. These parameter optimizations maintain high accuracy while improving processing efficiency.
Data Source
AI summary
Aspects of the disclosure relate to generating a pseudo-code from a text summarization based on a convolutional neural network. A computing platform may receive, by a computing device, a first document comprising text in a natural language different from English. Subsequently, the computing platform may translate, based on a neural machine translation model, the first document to a second document comprising text in English. Then, the computing platform may generate an attention-based convolutional neural network (CNN) for the second document. Then, the computing platform may extract, by applying the attention-based CNN, an abstractive summary of the second document. Subsequently, the computing platform may generate, based on the abstractive summary, a flowchart. Then, the computing platform may generate, based on the flowchart, a pseudo-code. Subsequently, the computing platform may display, via an interactive graphical user interface, the flowchart, and the pseudo-code.