{"__v":0,"_id":"5845a4a99f6fbb1b004307ef","category":{"version":"5845a4a89f6fbb1b004307b7","project":"54d3007669578e0d002730c9","_id":"5845a4a89f6fbb1b004307b9","__v":0,"sync":{"url":"","isSync":false},"reference":false,"createdAt":"2015-07-30T06:25:25.645Z","from_sync":false,"order":1,"slug":"key-concepts","title":"Key Concepts"},"parentDoc":null,"project":"54d3007669578e0d002730c9","user":"55bf6cdcad601c2b00762d13","version":{"__v":1,"_id":"5845a4a89f6fbb1b004307b7","project":"54d3007669578e0d002730c9","createdAt":"2016-12-05T17:32:24.708Z","releaseDate":"2016-12-05T17:32:24.708Z","categories":["5845a4a89f6fbb1b004307b8","5845a4a89f6fbb1b004307b9","5845a4a89f6fbb1b004307ba","5845a4a89f6fbb1b004307bb","5845a4a89f6fbb1b004307bc","5845a4a89f6fbb1b004307bd","5845a4a89f6fbb1b004307be","5845a4a89f6fbb1b004307bf","5845a4a89f6fbb1b004307c0"],"is_deprecated":false,"is_hidden":false,"is_beta":false,"is_stable":true,"codename":"","version_clean":"25.0.0","version":"25"},"updates":["5808ac04b2524d0f00350671"],"next":{"pages":[],"description":""},"createdAt":"2016-09-16T23:15:52.891Z","link_external":false,"link_url":"","githubsync":"","sync_unique":"","hidden":false,"api":{"results":{"codes":[]},"settings":"","auth":"required","params":[],"url":""},"isReference":false,"order":2,"body":"Machine Learning is a tool that allows your <a href=\"https://docs.api.ai/docs/concept-agents\" target=\"_blank\">agent</a> to understand user inputs in natural language and convert them into structured data, extracting relevant parameters. In the API.AI terminology, your agent uses machine learning algorithms to match user requests to specific <a href=\"https://docs.api.ai/docs/concept-intents\" target=\"_blank\">intents</a> and uses <a href=\"https://docs.api.ai/docs/concept-entities\" target=\"_blank\">entities</a> to extract relevant data from them.\n\nThe agent “learns” both from the data you provide in it (<a href=\"https://docs.api.ai/docs/concept-intents#section-example-and-template-modes\" target=\"_blank\">annotated examples</a> in intents and <a href=\"https://docs.api.ai/docs/concept-entities#entity-types\" target=\"_blank\">entries</a> in entities) and from the language models developed by API.AI. Based on this data, it builds a model (algorithm) for making decisions on which intent should be triggered by a user input and what data needs to be extracted. The model is unique to your agent.\n\nThe model adjusts dynamically according to the changes made in your agent and in the API.AI platform. To make sure that the model is improving, your agent needs to constantly be <a href=\"https://docs.api.ai/docs/training\" target=\"_blank\">trained</a> on real conversation logs.\n\nFor a quicker start of your agent’s development, you can use predefined knowledge packages – <a href=\"https://docs.api.ai/docs/domains\" target=\"_blank\">Domains</a>. We have created them for most requested use cases in our platform. And we continue improving them and adding new ones.\n\nThe machine learning model for your agent is updated every time you save changes in intents and entities, approve changes in Training, or <a href=\"https://docs.api.ai/docs/concept-agents#export-and-import\" target=\"_blank\">import/restore</a> an agent from a zip file.\n[block:callout]\n{\n  \"type\": \"warning\",\n  \"body\": \"For agents with more than 50 entities or more than 600 intents, you need to update the model manually. To do so, go to your agent settings > ML Settings and click the 'Train' button.\"\n}\n[/block]\nYou can think of an agent as a human in a certain way. Say, you have a new team member who has some knowledge already (similar to Domains for your agent). When you need to teach them something new, you start training them (similar to adding custom intents and entities). Machine learning allows the agent to make decisions based on the new data. It can make mistakes at first (same as humans). When you spend time on training, it starts making better decisions.","excerpt":"","slug":"machine-learning","type":"basic","title":"Machine Learning"}
Machine Learning is a tool that allows your <a href="https://docs.api.ai/docs/concept-agents" target="_blank">agent</a> to understand user inputs in natural language and convert them into structured data, extracting relevant parameters. In the API.AI terminology, your agent uses machine learning algorithms to match user requests to specific <a href="https://docs.api.ai/docs/concept-intents" target="_blank">intents</a> and uses <a href="https://docs.api.ai/docs/concept-entities" target="_blank">entities</a> to extract relevant data from them. The agent “learns” both from the data you provide in it (<a href="https://docs.api.ai/docs/concept-intents#section-example-and-template-modes" target="_blank">annotated examples</a> in intents and <a href="https://docs.api.ai/docs/concept-entities#entity-types" target="_blank">entries</a> in entities) and from the language models developed by API.AI. Based on this data, it builds a model (algorithm) for making decisions on which intent should be triggered by a user input and what data needs to be extracted. The model is unique to your agent. The model adjusts dynamically according to the changes made in your agent and in the API.AI platform. To make sure that the model is improving, your agent needs to constantly be <a href="https://docs.api.ai/docs/training" target="_blank">trained</a> on real conversation logs. For a quicker start of your agent’s development, you can use predefined knowledge packages – <a href="https://docs.api.ai/docs/domains" target="_blank">Domains</a>. We have created them for most requested use cases in our platform. And we continue improving them and adding new ones. The machine learning model for your agent is updated every time you save changes in intents and entities, approve changes in Training, or <a href="https://docs.api.ai/docs/concept-agents#export-and-import" target="_blank">import/restore</a> an agent from a zip file. [block:callout] { "type": "warning", "body": "For agents with more than 50 entities or more than 600 intents, you need to update the model manually. To do so, go to your agent settings > ML Settings and click the 'Train' button." } [/block] You can think of an agent as a human in a certain way. Say, you have a new team member who has some knowledge already (similar to Domains for your agent). When you need to teach them something new, you start training them (similar to adding custom intents and entities). Machine learning allows the agent to make decisions based on the new data. It can make mistakes at first (same as humans). When you spend time on training, it starts making better decisions.