人工智能培训

搜索

深度学习论文:自然语言与可解释AI模型的互动(Natural Language Interaction with Explainable AI Models)

[复制链接]
vix 发表于 2019-3-15 13:17:20 | 显示全部楼层 |阅读模式
vix 2019-3-15 13:17:20 279 0 显示全部楼层
深度学习论文:自然语言与可解释AI模型的互动(Natural Language Interaction with Explainable AI Models)本文介绍了一个可解释的AI(XAI)系统,该系统为其预测提供了解释。该系统由两个关键组件组成 - 即用于识别和定位输入数据中感兴趣概念的预测和/或图形(AOG)模型,以及用于向用户提供关于AOG预测的解释的XAI模型。在这项工作中,我们专注于指定的XAI模型,以自然语言与用户交互,而AOG的预测被认为是给定的,并由AOG的相应解析图(pg)表示。我们的XAI模型将pg作为输入,并使用以下类型的推理为用户的问题提供答案:directevidence(例如,检测得分),基于部分的推断(例如,检测到的部分为所提出的概念提供证据),以及来自时空的其他证据上下文(例如,来自时空环绕的约束)。我们使用Youtube Action数据集识别用户问题和XAI答案之间的几个相关性。
This paper presents an explainable AI (XAI) system that provides explanationsfor its predictions.The system consists of two key components -- namely, theprediction And-Or graph (AOG) model for recognizing and localizing concepts ofinterest in input data, and the XAI model for providing explanations to theuser about the AOG's predictions.In this work, we focus on the XAI modelspecified to interact with the user in natural language, whereas the AOG'spredictions are considered given and represented by the corresponding parsegraphs (pg's) of the AOG.Our XAI model takes pg's as input and providesanswers to the user's questions using the following types of reasoning: directevidence (eg, detection scores), part-based inference (eg, detected partsprovide evidence for the concept asked), and other evidences fromspatio-temporalcontext (eg, constraints from the spatio-temporal surround).We identify several correlations between user's questions and the XAI answersusing Youtube Action dataset.深度学习论文:自然语言与可解释AI模型的互动(Natural Language Interaction with Explainable AI Models) UNkU0Wp1u06KAsNp.jpg
URL地址:https://arxiv.org/abs/1903.05720     ----pdf下载地址:https://arxiv.org/pdf/1903.05720    ----深度学习论文:自然语言与可解释AI模型的互动(Natural Language Interaction with Explainable AI Models)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

vix当前离线
新手上路

查看:279 | 回复:0

快速回复 返回顶部 返回列表