基于迁移学习的基础测绘地址数据解析模型构建及应用

Construction and application of basic surveying address data parsing model based on transfer learning

摘要

基础测绘地址数据已被广泛使用于工业、农业和服务业,但此类数据库数据存在标准要素不统一的问题,影响地址数据的解析和应用.针对这一问题,构建了一种基于迁移学习的基础测绘地址数据解析模型.使用浙江省15548条基础测绘地址数据在已训练的MacBERT-base-Chinese普通地址解析模型上进行微调训练,并对模型进行了预测和效率评估.结果表明,该解析模型在解决基础测绘地址数据要素自动解析方面具有较好的效果,能够在较短的训练时间和有限的数据集上有效解析基础测绘地址数据.

Abstract

Basic surveying address data is widely used in industry, agriculture, and services. However, such databases often suffer from inconsistent standard elements, which hinders the parsing and application of address data. This study constructed a transfer learning-based model for parsing basic surveying address data. The model was fine-tuned on a pre-trained MacBERT-base-Chinese general address parsing model using 15548 basic surveying and mapping address data samples from Zhejiang Province. Subsequently, the prediction performance and efficiency of the model were evaluated. The results indicated that the proposed parsing model performs effectively in automatically parsing elements of basic surveying address data, achieving efficient parsing with limited training time and a constrained dataset.

关键词

基础测绘地址 / 地址解析 / 迁移学习 / MacBERT模型

Key words

basic surveying address / address parsing / transfer learning / MacBERT model

参考文献

[1] 张雪英, 闾国年, 李伯秋, 等. 基于规则的中文地址要素解析方法[J]. 地球信息科学, 2010(1): 9-16.

[2] 杨林, 余丽, 叶亚琴, 等. 一种中文门楼址的自适应表达及成分解析方法[J]. 测绘科学, 2014, 39(12): 109-115.

[3] 许也, 申柏希, 徐翔, 等. 基于条件随机场的非规范化中文地址解析方法[J]. 地理与地理信息科学, 2019, 35(2): 12-18.

[4] 吴恪涵, 张雪英, 叶鹏, 等. 基于BERT-BiLSTM-CRF的中文地址解析方法[J]. 地理与地理信息科学, 2021, 37(4): 10-15.

[5] 杨彬, 罗安, 王勇, 等. 一种顾及地理实体知识的中文地址要素解析方法[J]. 测绘科学, 2023, 48(9): 202-211.

[6] 余同瑞, 金冉, 韩晓臻, 等. 自然语言处理预训练模型的研究综述[J]. 计算机工程与应用, 2020, 56(23): 12-22.

[7] QIU X P, SUN T X, XU Y G, et al. Pre-trained models for natural language processing: a survey[J]. Science China Technological Sciences, 2020, 63(10): 1872-1897.

[8] 李舟军, 范宇, 吴贤杰. 面向自然语言处理的预训练技术研究综述[J]. 计算机科学, 2020, 47(3): 162-173.

[9] SUN Y, WANG S, LI Y, et al. ERNIE: enhanced representation through knowledge integration[PP/OL]. (2019-04-19)[2024-04-01]. https://arxiv.org/abs/1904.09223.

[10] CLARK K, LUONG M T, LE Q V, et al. ELECTRA: pre-training text encoders as discriminators rather than generators[PP/OL]. (2020-03-23)[2024-04-01]. https://arxiv.org/abs/2003.10555.

[11] DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[PP/OL]. (2019-06-01)[2024-04-01]. https://aclanthology.org/N19-1423.

[12] WEI J, REN X, LI X, et al. NEZHA: neural contextualized representation for Chinese language understanding[PP/OL]. (2019-08-31)[2024-04-01]. https://arxiv.org/abs/1909.00204.

[13] CUI Y M, CHE W X, LIU T, et al. Revisiting pre-trained models for Chinese natural language processing[PP/OL]. (2020-08-29)[2024-04-01]. https://arxiv.org/abs/2004.13922.

[14] 庄福振, 罗平, 何清, 等. 迁移学习研究进展[J]. 软件学报, 2015, 26(1): 26-39.

基金资助

杭州市地理信息公共服务平台运维服务项目(JCCHZF-2023-1007)