Concurrent Word and Sequence Labeling for Scientific Text Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional named entity recognition (NER) technologies are inaccurate when classifying unstructured text related to scientific topics, particularly due to the lack of standardized vocabulary and dictionaries, leading to failures in identifying hydrocarbon indicators and other scientific terms.
Innovation Solution
A computer-implemented text classification system that simultaneously identifies individual words and sequences of words pertaining to a topic, using a deep neural network model trained on labeled data with preprocessing techniques to extract and label sentences and paragraphs, thereby improving accuracy in scientific text classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional NER technologies are used to classify unstructured text, then the system can identify words using predefined dictionaries, but the accuracy deteriorates when dealing with scientific topics that lack standardized vocabulary
Solution Approach 1:
The model is designed to perform multiple classification tasks simultaneously - both word-level classification and sequence-level classification - making it universally applicable to different text types and vocabularies, including scientific domains without standardized dictionaries
Solution Approach 2:
The patent introduces a new dimension of sequence-level classification alongside traditional word-level classification. This dual-level approach allows the system to capture contextual information at the sequence level while maintaining word-level precision, thereby improving overall classification accuracy for scientific texts
2Measurement precision
If the model performs only word-level classification, then the computational complexity is lower, but the classification accuracy for scientific topics deteriorates
Solution Approach 1:
The patent merges word-level classification and sequence-level classification into a single unified model. This combination allows the system to leverage both fine-grained word information and broader sequence context, improving classification accuracy for scientific topics while maintaining computational efficiency through shared model components
3Ease of manufacture
If conventional NER systems rely on predefined dictionaries, then the system structure is simpler, but the system fails to identify scientific terms without standardized vocabulary
Solution Approach 1:
The model is trained to learn vocabulary and classification patterns directly from training data without relying on predefined dictionaries. This self-learning capability allows the system to automatically adapt to scientific domains and identify terms even when no standardized vocabulary exists, while maintaining relatively simple implementation through standard deep learning techniques
Data Source
AI summary
A computing system includes a processor and memory that stores instructions that, when executed by the processor, cause the processor to perform several acts. The acts include providing tokens as input to a computer-implemented model, where the tokens are representative of a sequence of words, and further where the computer-implemented model has been trained to identify sets of tokens that pertain to a topic and individual tokens within the sets of tokens that pertain to the topic. The acts also include obtaining, from the computer-implemented model: 1) a first label assigned to a token within the tokens by the computer-implemented model, where the first label indicates that a word represented by the token pertains to the topic; and 2) a second label assigned collectively to the tokens by the computer-implemented model, where the second label indicates that the sequence of words represented by the tokens collectively pertains to the topic.