Inter-sentential Relation Extraction via Dependency Parse Tree Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing relation extraction methods primarily focus on binary relations within single sentences, ignoring inter-sentential relationships, which results in loss of information and limited knowledge extraction from documents.
Innovation Solution
A computer-implemented method generates dependency parse trees for relevant sentences, links root nodes to create a chain of words, generates subtree embedding vectors, and uses a recurrent neural network to classify semantic relationships between entities across sentences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If relation extraction is performed only within single sentences (intra-sentential), then the method complexity remains manageable and processing speed is maintained, but inter-sentential relationships are lost and information completeness deteriorates
Solution Approach 1:
The patent segments the document into individual sentences and generates separate dependency parse trees for each sentence. This segmentation allows the system to handle complex inter-sentential relationships by breaking them down into manageable sentence-level units while still capturing cross-sentence connections through the neural network architecture.
Solution Approach 2:
The patent introduces an intermediary mechanism using recurrent neural networks that can process and maintain information across sentence boundaries. The RNN acts as a mediator that connects intra-sentential dependency structures with inter-sentential relationship extraction, enabling the system to capture relationships across sentences without overwhelming complexity.
2Measurement precision
If dependency parse trees are generated for all sentences and linked to capture inter-sentential relationships, then relationship extraction accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary action by generating dependency parse trees for all sentences in advance before conducting the actual relationship extraction. This preprocessing step organizes the textual information into structured representations that facilitate more efficient subsequent processing and relationship identification across sentences.
Solution Approach 2:
The patent employs dynamic recurrent neural networks that can adaptively process variable-length sequences of dependency parse trees. The dynamic nature of the RNN allows it to efficiently handle varying document lengths and relationship complexities, adjusting its processing to balance accuracy with computational efficiency.
3Productivity
If traditional binary relation extraction methods are used within single sentences, then the extraction process is simple and fast, but the ability to capture semantic relationships across multiple sentences is limited
Solution Approach 1:
The patent implements a universal extraction framework that handles both intra-sentential and inter-sentential relationships using the same neural network architecture. This multi-functional approach allows the system to perform simple binary relation extraction within sentences while simultaneously capturing complex relationships across multiple sentences, maximizing knowledge extraction capability.
Solution Approach 2:
The patent changes the parameter of relationship scope from sentence-level to document-level by modifying the input structure to the neural network. By feeding sequences of dependency parse trees into the RNN, the system transforms the extraction task to operate at the document level while maintaining the computational tractability through parameter optimization.
Data Source
AI summary
Provided is a computer-implemented method for inter-sententially determining a semantic relationship between a first entity and a second entity in a natural language document, comprising at least the steps of:generating a first dependency parse tree, DPT, for a first origin sentence of the document which comprises the first entity, wherein each DPT comprises at least a root node;generating a second DPT for a second origin sentence of the document which mentions the second entity;linking the root nodes of the first DPT and the second DPT so as to create a chain of words, COW;determining for each word in the COW a subtree;generating for each word in the COW a subtree embedding vector cw which is based at least on word embedding vectors xw of the words of the subtree;generating a representation vector pw for each word in the COW; andclassifying, using a recurrent neural network, the semantic relationship between the first entity and the second entity, based on the input representation vectors pw.