Evaluating the Performance of Interpretability Methods in Text Categorization Task

IF 0.8 Q2 MATHEMATICS

Lobachevskii Journal of Mathematics Pub Date : 2024-07-19 DOI:10.1134/s1995080224600699

A. A. Rogov, N. V. Loukachevitch

引用次数: 0

Abstract

Neural networks are progressively assuming a larger role in individuals daily routines, as their complexity continues to grow. While the model demonstrates satisfactory performance when evaluated on the test data, it often yields unforeseen outcomes in real-world scenarios. To diagnose the source of these errors, understanding the decision-making process employed by the model becomes crucial. In this paper, we consider various methods of interpreting the BERT model in classification tasks, and also consider methods for evaluating interpretation methods using vector representations fastText, GloVe and Sentence-BERT.

Abstract Image

查看原文本刊更多论文

评估文本分类任务中可解释性方法的性能

摘要随着神经网络复杂性的不断提高，它在个人日常工作中正逐步发挥着越来越大的作用。虽然模型在测试数据中表现出令人满意的性能，但在现实世界中却经常产生不可预见的结果。要诊断这些错误的根源，了解模型所采用的决策过程至关重要。在本文中，我们考虑了在分类任务中解释 BERT 模型的各种方法，还考虑了使用向量表示 fastText、GloVe 和 Sentence-BERT 评估解释方法的方法。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Lobachevskii Journal of Mathematics MATHEMATICS-

CiteScore

1.50

自引率

42.90%

发文量

127

期刊介绍： Lobachevskii Journal of Mathematics is an international peer reviewed journal published in collaboration with the Russian Academy of Sciences and Kazan Federal University. The journal covers mathematical topics associated with the name of famous Russian mathematician Nikolai Lobachevsky (Lobachevskii). The journal publishes research articles on geometry and topology, algebra, complex analysis, functional analysis, differential equations and mathematical physics, probability theory and stochastic processes, computational mathematics, mathematical modeling, numerical methods and program complexes, computer science, optimal control, and theory of algorithms as well as applied mathematics. The journal welcomes manuscripts from all countries in the English language.