Three-Layer Retrieval and Self-Evaluation Classification Method Based on FastText Algorithm
Patent big data serves as a valuable scientific research source for technological innovation, enabling breakthroughs beyond existing knowledge and fostering disruptive ideas. One key challenge in this field is how to efficiently obtain patent documents quickly and accurately. This is a critical focu...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11003909/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Patent big data serves as a valuable scientific research source for technological innovation, enabling breakthroughs beyond existing knowledge and fostering disruptive ideas. One key challenge in this field is how to efficiently obtain patent documents quickly and accurately. This is a critical focus in the exploration of patent search methodologies. Our approach differs from conventional patent search processes. We have developed a three-level patent classification method that utilizes a multi-step search strategy with specific constraints, alongside an innovative classification system based on the FastText algorithm. By combining these techniques with an emphasis on recall ratio, we can test the efficacy of each level of the database boundaries. This enables swift identification of target patents and allows for focused screening in specific fields, providing robust support for technical or product innovation activities. Furthermore, we applied this method to the retrieval of DBDI patent data, which represents one of the three leading commercial direct ionization ion source technologies globally. Our classification results indicate a remarkable accuracy of 96.97%, reflecting a 21.97% improvement over the TextRNN_Att text algorithm. This effectively demonstrates the success of our proposed methodology. Overall, this study offers a theoretical framework for researching multi-level classification methods in logical retrieval and provides a practical foundation for classifying direct ionization and ionization technologies. |
|---|---|
| ISSN: | 2169-3536 |