基于机器学习的图书采选自动化辅助——以西雅图公共图书馆开放数据为例

Computer-Aided Book Selection and Acquisitions with Machine Learning:A Case Study of Open Data from Seattle Public Library

构建能够通过图书标题自动化预测图书外借概率的机器学习模型,以便作为图书采访选书时的辅助标准。当选书清单中图书数量很大时,自动化工具辅助非常有帮助。使用词嵌入实现自然语言文本的向量化表示。构建了人工神经网络模型,并用西雅图公共图书馆开放数据中2017年度的外借记录对模型进行了训练。调试超参数以获得更优化的结果。模型在测试集上准确率高于77%,且在两个分类上均具有可接受的查全率和查准率。模型对书籍外借可能性的预测能辅助图书采选。

A machine learning model is built to automatically predict a book being borrowed based on its title.The model can be used to aid book selection in the acquisition,especially when there are a large number of books in the book list.Word embeddings are used to represent texts in natural language as vectors.The author of this paper then builds a supervised learning model with artificial neural networks(ANN).Checkout records of the year 2017 from Seattle Public Library's open data are used to train the model.The model achieves an accuracy of over 77%on the testing set and presents acceptable recalls and precisions on both classes.This suggests that the model is helpful for selecting books in library acquisitions.