通过评审终结点话语启用主动学习的相关概念Concepts for enabling active learning by reviewing endpoint utterances
主动学习是提高预测准确性的三个策略之一，也是最易于实现的策略。Active learning is one of three strategies to improve prediction accuracy and the easiest to implement. 评审终结点话语通过主动学习获取正确的意向和实体。With active learning, your review endpoint utterances for correct intent and entity. LUIS 选择它不确定的终结点话语。LUIS chooses endpoint utterances it is unsure of.
什么是主动学习What is active learning
主动学习是一个两步过程。Active learning is a two-step process. 首先，LUIS 选择其在应用终结点收到的需要验证的陈述。First, LUIS selects utterances it receives at the app's endpoint that need validation. 评审陈述后，再次训练并发布应用。After reviewing the utterances, train and publish the app again.
评审列表上有哪些陈述Which utterances are on the review list
首要触发意向分数较低或者两个最高的意向分数过于接近时，LUIS 会将陈述添加到评审列表。LUIS adds utterances to the review list when the top firing intent has a low score or the top two intents' scores are too close.
每个应用的话语的单个池Single pool for utterances per app
“查看终结点话语”列表不会根据版本更改。The Review endpoint utterances list doesn't change based on the version. 要审核的话语只有一个池，不管经常编辑哪个版本的话语，也不管在终结点上发布哪个版本的应用。There is a single pool of utterances to review, regardless of which version the utterance you are actively editing or which version of the app was published at the endpoint.
陈述来自哪里Where are the utterances from
终结点陈述来自应用程序 HTTP 终结点上的最终用户查询。Endpoint utterances are taken from end-user queries on the application’s HTTP endpoint. 如果未发布应用或应用未收到点击，则没有要评审的陈述。If your app is not published or has not received hits yet, you do not have any utterances to review. 如果未收到针对特定意向或实体的终结点点击，则没有包含它们的陈述需要进行评审。If no endpoint hits are received for a specific intent or entity, you do not have utterances to review that contain them.
定期计划评审Schedule review periodically
不需要每天评审建议的陈述，但应将此纳入 LUIS 的常规维护。Reviewing suggested utterances doesn't need to be done every day but should be part of your regular maintenance of LUIS.