Document-Graph Q&A Boosting for Scientific Instrument Identifiers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing generative artificial intelligence techniques for answering questions about scientific instruments suffer from insufficient accuracy due to the challenges posed by scientific instrument identifiers, which often carry little semantic meaning and can be affected by minor spelling differences, leading to incorrect text block selection and inaccurate answers.
Innovation Solution
Implementing a document-graph repository with context-tagged text blocks, iterative graph-walking and embedding-change comparison, text block re-ranking based on chain-of-thought prompts, and prompt augmentation for identifier emphasis to enhance answer accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing generative AI techniques are used for answering questions aboutscientific instruments, then the system is simple and easy to operate, but the answer accuracy is insufficient due to challenges withscientific instrument identifiers
Solution Approach 1:
The patent segments the text blocks by inserting separator tokens between different text blocks and their associated instrument identifiers. This segmentation allows the model to better distinguish between text content and identifier information, improving the accuracy of matching questions to relevant text blocks while maintaining manageable system complexity through structured data organization.
Solution Approach 2:
The patent introduces an intermediary processing layer that handles scientific instrument identifiers separately from the main text content. This intermediary mechanism processes identifiers through special tokenization and embedding methods, acting as a mediator between the raw input data and the generative AI model, thereby improving answer accuracy without requiring complete system redesign.
2Measurement precision
If standard text block selection methods are used, then the process is simple and fast, but spelling differences in instrument identifiers lead to incorrect text block selection
Solution Approach 1:
The patent applies preliminary action by pre-processing text blocks to insert separator tokens and standardize identifier representations before the actual question-answering process. This advance preparation ensures that spelling variations in instrument identifiers are normalized, enabling more accurate text block selection without adding significant processing time during query execution.
Solution Approach 2:
The patent changes the parameter representation of instrument identifiers by transforming them into standardized embeddings with separator tokens. This parameter transformation allows the system to match identifiers despite spelling differences, improving text block selection accuracy while maintaining efficient processing through optimized embedding comparisons.
Data Source
AI summary
Systems or techniques are provided for facilitating retrieval augmented generative question and answer boosting. In various embodiments, a system can access a plain text question regarding a scientific instrument. In various aspects, the system can generate, via a large language model that references a document-graph repository, a structured or unstructured answer for the plain text question. In various instances, the document-graph repository can comprise a plurality of document-graphs that respectively correspond to a plurality of technical documents. In various cases, for a first document-graph that corresponds to a first technical document, leaf nodes of the first document-graph can represent respective text blocks written in the first technical document, and non-leaf nodes of the first document-graph can respectively represent a document title, one or more section headings, and one or more scientific instrument identifiers written in the first technical document and beneath which the respective text blocks are nested.