Concurrent Graph and Text Annotation Editor for Entity Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current annotation editors for question answer systems are inefficient as they do not provide a comprehensive view of entities and relations throughout a document, making it difficult for human annotators to identify co-references and relations, especially when entities are not close together, leading to potential missed annotations.
Innovation Solution
Concurrently displaying a graphical editor and a text editor allows annotators to view and interact with entities and relations graphically, repositioning nodes to facilitate easy identification and annotation of co-references and relations across the document.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human annotators scroll through the document to identify entities for linking, then they can potentially find all entities, but the process becomes time-consuming and error-prone
Solution Approach 1:
The patent transforms the one-dimensional linear document structure into a two-dimensional graphical representation where entities are displayed as nodes and relations as edges. This dimensional change allows annotators to perceive entity relationships and document structure simultaneously, eliminating the need to scroll through the entire document linearly while maintaining complete annotation coverage.
2Ease of operation
If today's annotation editors display only linear text, then the interface remains simple, but annotators cannot efficiently identify co-references and relations between entities
Solution Approach 1:
The patent segments the document into discrete entity nodes and relation edges, allowing annotators to interact with individual entities and their relationships independently. This segmentation enables efficient identification of co-references and relations by displaying only relevant portions of the graph, while the overall interface remains manageable through hierarchical organization.
Solution Approach 2:
The patent implements a nested graphical interface where detailed entity information and related entities are organized in hierarchical layers. The main graph display shows high-level entity relationships, while detailed views of individual entities can be accessed through nesting, allowing the interface to adapt its complexity based on the current view without overwhelming the user.
3Loss of information
If entities are displayed in their original document positions, then the graphical editor reflects document structure, but entities far apart in the document remain difficult to connect
Solution Approach 1:
The patent implements a dynamic graphical editor where node positions can be automatically adjusted based on entity relationships and document structure. The system dynamically repositions entities to group related items together while maintaining links to their original document locations, allowing annotators to easily identify relations between distant entities without losing document structure context.
Data Source
AI summary
An approach is provided in which the approach displays a graphical editor and a text editor concurrently on a display. The text editor displays a document text that includes multiple entities and the graphical editor displays multiple nodes corresponding to the multiple entities. In response to receiving a user selection on the graphical editor that selects a first one of the multiple nodes, the approach identifies, based on the document text, one or more second nodes that correspond to the first node and repositions the one or more second nodes to a second position on the graphical editor based on a first position of the first node.