Scholarlink

<p>The major objective of patent retrieval is to identify relevant documents from a large database. Patent retrieval requests often take the form of patent claims, as in prior art searches, and these searches typically require a long time to complete. However, the significant discrepancy between the query and the relevant documents leads to poor retrieval efficiency. Past studies have attempted to resolve these mismatch issues by applying query expansion approaches, which are usually successful in various retrieval tasks. One of the primary challenges in patent retrieval is vocabulary mismatch, which complicates the retrieval process. Nevertheless, it has been found that query expansion methods produce poor results in patent searches. To overcome this challenge, this research proposes an effective patent retrieval approach utilizing query expansion. Here, query expansion is applied to retrieve patents from the provided data. The necessary textual data are collected from multiple sources. The acquired data are then passed through a text pre-processing phase. The pre-processed data are further input to the feature extraction stage, where “Bidirectional Encoder Representations from Transformers (BERT)” and “Generative Pretrained Transformer-3 (GPT-3)” models are used to extract features. Moreover, the features obtained from BERT and GPT-3 are concatenated and fed into an Optimal Bi-Clustering (OBi-C) algorithm for patent retrieval, with parameters tuned using the Enhanced Social Engineering Optimizer (ESEO) to improve performance. Finally, experimental analysis is conducted on the developed patent retrieval system to validate its effectiveness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Intelligent Feature Concatenation Process-Based Effective Query Expansion for Patent Retrieval Approach Using Optimal Bi-clustering and Enhanced Social Engineering Optimizer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月10日

查看全文
  • G. David Raj,
  • Saswati Mukherjee,
  • C. R. Rene Robin,
  • R. L. Jasmine

摘要

The major objective of patent retrieval is to identify relevant documents from a large database. Patent retrieval requests often take the form of patent claims, as in prior art searches, and these searches typically require a long time to complete. However, the significant discrepancy between the query and the relevant documents leads to poor retrieval efficiency. Past studies have attempted to resolve these mismatch issues by applying query expansion approaches, which are usually successful in various retrieval tasks. One of the primary challenges in patent retrieval is vocabulary mismatch, which complicates the retrieval process. Nevertheless, it has been found that query expansion methods produce poor results in patent searches. To overcome this challenge, this research proposes an effective patent retrieval approach utilizing query expansion. Here, query expansion is applied to retrieve patents from the provided data. The necessary textual data are collected from multiple sources. The acquired data are then passed through a text pre-processing phase. The pre-processed data are further input to the feature extraction stage, where “Bidirectional Encoder Representations from Transformers (BERT)” and “Generative Pretrained Transformer-3 (GPT-3)” models are used to extract features. Moreover, the features obtained from BERT and GPT-3 are concatenated and fed into an Optimal Bi-Clustering (OBi-C) algorithm for patent retrieval, with parameters tuned using the Enhanced Social Engineering Optimizer (ESEO) to improve performance. Finally, experimental analysis is conducted on the developed patent retrieval system to validate its effectiveness.