Oddballness:利用语言模型进行通用异常检测
Oddballness: universal anomaly detection with language models
AI总结:
本文提出一种基于语言模型的无监督异常检测度量 oddballness,用于衡量词元的“奇怪”程度,并在语法错误检测任务中证明其优于仅依赖低似然事件的方法。
摘要:
我们提出了一种新方法,以完全无监督的方式利用语言模型检测文本(一般指任意数据序列)中的异常。该方法考虑语言模型生成的概率(似然),但不关注低似然词元,而是考虑本文引入的新度量:oddballness。Oddballness 衡量给定词元在语言模型看来有多“奇怪”。我们在语法错误检测任务(文本异常检测的一个特例)中证明,在假设完全无监督设置的情况下,oddballness 优于仅考虑低似然事件的方法。
英文摘要:
We present a new method to detect anomalies in texts (in general: in sequences of any data), using language models, in a totally unsupervised manner. The method considers probabilities (likelihoods) generated by a language model, but instead of focusing on low-likelihood tokens, it considers a new metric introduced in this paper: oddballness. Oddballness measures how ``strange'' a given token is according to the language model. We demonstrate in grammatical error detection tasks (a specific case of text anomaly detection) that oddballness is better than just considering low-likelihood events, if a totally unsupervised setup is assumed.
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。