融合預(yù)訓(xùn)練語言模型的知識圖譜在政務(wù)問答系統(tǒng)中的應(yīng)用研究

打開文本圖片集

摘要：該文針對當(dāng)前政務(wù)問答系統(tǒng)面臨的復(fù)雜語境理解、政策法規(guī)解釋等問題，探討了如何將預(yù)訓(xùn)練語言模型與知識圖譜進(jìn)行有效融合，以實現(xiàn)更加精準(zhǔn)、全面和個性化的政務(wù)信息問答服務(wù)，構(gòu)建了政務(wù)問答系統(tǒng)框架，利用知識圖譜和大模型工具驗證了該方法在提高問答準(zhǔn)確率、增強上下文理解能力方面的顯著優(yōu)勢。

關(guān)鍵詞：知識圖譜；自然語言處理；預(yù)訓(xùn)練語言模型；三元組；知識庫

doi：10.3969/J.ISSN.1672-7274.2024.09.063

中圖分類號：TP 3                 文獻(xiàn)標(biāo)志碼：A            文章編碼：1672-7274（2024）09-0-03

Research on the Application of Knowledge Graph Integrated with Pre-trained Language Models in Government Question-answering Systems

ZHANG Chaoyang， SHEN Jianhui， YE Weirong

（Zhejiang Public Information Industry Co.， LTD.， Hangzhou 310000， China）

Abstract： Aiming at the problems of complex context understanding and interpretation of policies and regulations faced by the current government question answering system， this paper discusses how to effectively integrate pre-trained language models and knowledge graphs， so as to realize more accurate， comprehensive and personalized government information question answering service. The framework of government question answering system is constructed， and the significant advantages of this method in improving the accuracy of question answering and enhancing the context understanding ability are verified by using knowledge graph and large model tools.

Keywords： knowledge graph; natural language processing; pre-trained language model; triple; knowledge base

0   引言

政務(wù)問答系統(tǒng)的核心在于如何更好地建模語言、理解和輸出文本信息，本文以政務(wù)服務(wù)垂直領(lǐng)域在線咨詢問答場景為例，探索預(yù)訓(xùn)練語言模型與知識圖譜的融合應(yīng)用。（剩余3992字）