简体繁体 English

路易斯的意图如何发挥作用？

[英]How does the intent in luis work?

原文 2018-06-04 06:35:51 3 1 c#/ botframework/ luis

How does it actually work for the intent ? 它实际上如何达到目的？ cause in my program i have more than 5 intents and it just shows if i test "hello" it say 0.20 in welcome while the other is just 0 and some have 0.01 or other and adding all of them up doesn't reach 1.0, hence it goes to the welcome intent which is what i expected. 原因在我的程序中，我有5个以上的意图，它只是显示我是否在测试“ hello”，它表示欢迎0.20，而另一个仅是0，其中一些具有0.01或其他，而所有这些加起来并没有达到1.0，因此这符合我的期望。

However how does this intent actually work, from the example in the website i was under assumption this is based on percentage based with 1 as the biggest value, just like a logistic regression. 但是，此意图实际上是如何工作的，根据我在网站上的示例得出的假设，此假设基于以1为最大值的百分比为基础，就像logistic回归一样。 However this doesn't seem to be the case since the total of intent value are less than 1.0 so i was thinking this number might be a confidence value or this is just totally different number than what i thought it was ? 但是，由于意图值的总和小于1.0，因此似乎并非如此，因此我认为此数字可能是置信度值，或者这与我原本的想法完全不同？ please help 请帮忙

1 个解决方案

Once you train your LUIS app and test it with an utterance it gives you the confidence level against each trained intent.This score you see is the degree of confidence LUIS has for that result. 一旦您对LUIS应用程序进行了培训并进行了语音测试，它就会为您提供针对每种受过培训的意图的置信度。您所看到的分数就是LUIS对于该结果的置信度。

To improve the score for an intent you should train your LUIS app with more utterances so that it can differentiate it among other intents. 为了提高意图的分数，您应该用更多的话语训练您的LUIS应用程序，以便可以将其与其他意图区分开。

You can also check out this paper by Microsof to understand how LUIS works internally. 您也可以通过Microsof查看本白皮书，以了解LUIS在内部如何工作。