Document Creation Support Apparatus for Medical Image Text Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatically generated texts based on medical images may lack important information or include non-relevant details, leading to a mismatch with user requests.
Innovation Solution
A document creation support apparatus that generates a variety of texts with different descriptions and expressions for feature portions in medical images, allowing users to select the most relevant text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single automatically generated text is produced based on medical images, then the document creation process is efficient and quick, but the text may miss important information or include non-important information, leading to a mismatch with user requests
Solution Approach 1:
The system changes the parameter of text generation from single-output to multi-output mode. By configuring the text generation unit to create multiple candidate texts with different description depths and focuses, the system maintains efficiency while improving reliability. Users can select from multiple texts with varying levels of detail and perspectives, ensuring at least one text accurately matches their specific needs.
Solution Approach 2:
The system introduces dynamic text generation where the number, type, and characteristics of generated texts can be adjusted based on user requirements. The text generation unit dynamically creates texts with different properties (e.g., some with detailed descriptions, others with concise summaries), allowing the system to adapt to various user preferences while maintaining high productivity.
2Reliability
If multiple candidate texts are generated with different descriptions, then the likelihood of including user-requested information increases, but the complexity of the system increases
Solution Approach 1:
The text generation unit is segmented into multiple specialized sub-units, each responsible for generating texts with specific characteristics (e.g., detailed descriptions, concise summaries, different focal points). This segmentation allows the system to produce diverse candidate texts without requiring a single complex generation model, thereby managing system complexity while improving text accuracy and relevance.
Solution Approach 2:
The text generation unit is designed with multi-functionality, capable of generating various types of texts (different lengths, styles, and focuses) using a unified architecture. This universal design avoids the need for multiple separate systems, reducing overall system complexity while still providing diverse text options that improve reliability.
3Reliability
If multiple texts with diverse content and expressions are generated, then user request satisfaction increases, but the processing time and computational resources increase
Solution Approach 1:
The system implements partial action by generating a limited number of high-quality candidate texts (e.g., 3-5 texts) rather than exhaustively generating all possible variations. This partial generation approach provides sufficient diversity to satisfy user requests while avoiding the excessive processing time and computational resources that would result from generating an unlimited number of texts.
Solution Approach 2:
The system optimizes generation parameters such as temperature, top-k, and top-p values to control the diversity and quality of generated texts. By carefully tuning these parameters, the system produces sufficiently diverse candidate texts that match user requests while maintaining efficient processing speeds and reasonable computational resource usage.
Data Source
AI summary
A text generation unit (14) generates a plurality of texts which describe properties of feature portions and are different from each other for at least one feature portion included in an image. A display control unit (15) performs control such that each of the plurality of texts is displayed on a display unit.