arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.05018cs.IRcs.AI
收藏

在嵌入草堆中寻找针:基于 Bagging 与 SVR 集成的法律文档检索

Finding Needles in Emb(a)dding Haystacks: Legal Document Retrieval via Bagging and SVR Ensembles

Goethe University Frankfurt(法兰克福大学)

Kevin Bönisch, Alexander Mehler

AI总结:

本文面向 GerDaLIR 法律检索,将任务拆解为多个二元“大海捞针”子问题,并用 bagging 与 SVR 投票集成在嵌入空间中检索,在不训练深度学习模型的情况下提升召回率。

摘要:

我们在 German Dataset for Legal Information Retrieval(GerDaLIR,德国法律信息检索数据集)上提出一种利用 Support Vector Regression(SVR,支持向量回归)集成、bootstrap aggregation(bagging,自助聚合)和嵌入空间的检索方法。通过将检索任务概念化为多个二元“大海捞针”子任务,我们证明,投票集成相较基线取得了更高的召回率(0.849 > 0.803 | 0.829),且未训练或微调任何深度学习模型,显示出有前景的初步结果。该方法具有进一步提升的潜力,尤其是通过改进编码模型和优化超参数。

英文摘要:

We introduce a retrieval approach leveraging Support Vector Regression (SVR) ensembles, bootstrap aggregation (bagging), and embedding spaces on the German Dataset for Legal Information Retrieval (GerDaLIR). By conceptualizing the retrieval task in terms of multiple binary needle-in-a-haystack subtasks, we show improved recall over the baselines (0.849 > 0.803 | 0.829) using our voting ensemble, suggesting promising initial results, without training or fine-tuning any deep learning models. Our approach holds potential for further enhancement, particularly through refining the encoding models and optimizing hyperparameters.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑