LUIS 应用的企业策略Enterprise strategies for a LUIS app

查看企业应用的设计策略。Review these design strategies for your enterprise app.

LUIS 请求超出配额When you expect LUIS requests beyond the quota

LUIS 基于 Azure 资源的定价层,具有每月配额和每秒配额。LUIS has a monthly quota as well as a per second quota, based on the pricing tier of the Azure resource.

如果 LUIS 应用请求速率超过了允许的配额速率,你可以:If your LUIS app request rate exceeds the allowed quota rate, you can:

使用多个具有相同应用定义的应用Use multiple apps with same app definition

导出原始 LUIS 应用,然后将该应用导入回单独的应用。Export the original LUIS app, then import the app back into separate apps. 每个应用都有其自己的应用 ID。Each app has its own app ID. 在发布时,请为每个应用创建单独的密钥,而不是对所有应用使用相同的密钥。When you publish, instead of using the same key across all apps, create a separate key for each app.

若要在所有应用之间获取同样的最高意向,请确保第一意向和第二意向之间的差距足够大,不会让 LUIS 混淆,针对陈述中微小的变化在应用间给出不同的结果。In order to get the same top intent between all the apps, make sure the intent prediction between the first and second intent is wide enough that LUIS is not confused, giving different results between apps for minor variations in utterances.

训练这些同级应用时,请确保使用所有数据进行训练When training these sibling apps, make sure to train with all data.

将单个应用指定为主应用。Designate a single app as the master. 建议查看的任何陈述都应添加到主应用,然后移回所有其他应用。Any utterances that are suggested for review should be added to the master app then moved back to all the other apps. 这是应用的一次完整导出,或是将主应用中已标记的陈述加载到子级。This is either a full export of the app, or loading the labeled utterances from the master to the children. 可从 LUIS 网站或者单个话语批量话语的创作 API 完成加载。Loading can be done from either the LUIS website or the authoring API for a single utterance or for a batch.

计划定期评审(例如每两周一次)终结点话语以进行主动学习,然后重新训练并重新发布。Schedule a periodic review, such as every two weeks, of endpoint utterances for active learning, then retrain and republish.

将多个 LUIS 密钥分配到相同的应用Assign multiple LUIS keys to same app

如果 LUIS 应用接收到的终结点命中数超过了单个密钥的配额,请创建更多密钥并将它们分配到该 LUIS 应用。If your LUIS app receives more endpoint hits than your single key's quota allows, create and assign more keys to the LUIS app. 创建流量管理器或负载均衡器,管理这些终结点密钥间的终结点查询。Create a traffic manager or load balancer to manage the endpoint queries across the endpoint keys.

单体应用返回错误意向When your monolithic app returns wrong intent

如果应用要预测多种陈述,请考虑实施调度模型If your app is meant to predict a wide variety of user utterances, consider implementing the dispatch model. 分解单体应用可让 LUIS 成功地专注于意向间的检测,而不会在父应用和子应用中的意向间产生混淆。Breaking up a monolithic app allows LUIS to focus detection between intents successfully instead of getting confused between intents across the parent app and child apps.

计划定期(例如每两周一次)查看终结点陈述以进行主动学习,然后重新训练并重新发布。Schedule a periodic review of endpoint utterances for active learning, such as every two weeks, then retrain and republish.

调度模型中的意向限制Intent limits in dispatch model

一个调度应用程序最多可包含 500 个调度资源,相当于 500 个意向。A dispatch application has 500 dispatch sources, equivalent to 500 intents, as the maximum.

详细信息More information

后续步骤Next steps