RecognizeUtterance - Amazon Lex

RecognizeUtterance

Sends user input to Amazon Lex V2. You can send text or speech. Clients use this API to send text and audio requests to Amazon Lex V2 at runtime. Amazon Lex V2 interprets the user input using the machine learning model built for the bot.

The following request fields must be compressed with gzip and then base64 encoded before you send them to Amazon Lex V2.

  • requestAttributes

  • sessionState

The following response fields are compressed using gzip and then base64 encoded by Amazon Lex V2. Before you can use these fields, you must decode and decompress them.

  • inputTranscript

  • interpretations

  • messages

  • requestAttributes

  • sessionState

The example contains a Java application that compresses and encodes a Java object to send to Amazon Lex V2, and a second that decodes and decompresses a response from Amazon Lex V2.

Request Syntax

POST /bots/botId/botAliases/botAliasId/botLocales/localeId/sessions/sessionId/utterance HTTP/1.1 x-amz-lex-session-state: sessionState x-amz-lex-request-attributes: requestAttributes Content-Type: requestContentType Response-Content-Type: responseContentType inputStream

URI Request Parameters

The request uses the following URI parameters.

botAliasId

The alias identifier in use for the bot that should receive the request.

Required: Yes

botId

The identifier of the bot that should receive the request.

Length Constraints: Fixed length of 10.

Pattern: ^[0-9a-zA-Z]+$

Required: Yes

localeId

The locale where the session is in use.

Length Constraints: Minimum length of 1.

Required: Yes

requestAttributes

Request-specific information passed between the client application and Amazon Lex V2

The namespace x-amz-lex: is reserved for special attributes. Don't create any request attributes for prefix x-amz-lex:.

The requestAttributes field must be compressed using gzip and then base64 encoded before sending to Amazon Lex V2.

requestContentType

Indicates the format for audio input or that the content is text. The header must start with one of the following prefixes:

  • PCM format, audio data must be in little-endian byte order.

    • audio/l16; rate=16000; channels=1

    • audio/x-l16; sample-rate=16000; channel-count=1

    • audio/lpcm; sample-rate=8000; sample-size-bits=16; channel-count=1; is-big-endian=false

  • Opus format

    • audio/x-cbr-opus-with-preamble;preamble-size=0;bit-rate=256000;frame-size-milliseconds=4

  • Text format

    • text/plain; charset=utf-8

Length Constraints: Minimum length of 1.

Required: Yes

responseContentType

The message that Amazon Lex V2 returns in the response can be either text or speech based on the responseContentType value.

  • If the value is text/plain;charset=utf-8, Amazon Lex V2 returns text in the response.

  • If the value begins with audio/, Amazon Lex V2 returns speech in the response. Amazon Lex V2 uses Amazon Polly to generate the speech using the configuration that you specified in the requestContentType parameter. For example, if you specify audio/mpeg as the value, Amazon Lex V2 returns speech in the MPEG format.

  • If the value is audio/pcm, the speech returned is audio/pcm at 16 KHz in 16-bit, little-endian format.

  • The following are the accepted values:

    • audio/mpeg

    • audio/ogg

    • audio/pcm (16 KHz)

    • audio/* (defaults to mpeg)

    • text/plain; charset=utf-8

Length Constraints: Minimum length of 1.

sessionId

The identifier of the session in use.

Length Constraints: Minimum length of 2. Maximum length of 100.

Pattern: [0-9a-zA-Z._:-]+

Required: Yes

sessionState

Sets the state of the session with the user. You can use this to set the current intent, attributes, context, and dialog action. Use the dialog action to determine the next step that Amazon Lex V2 should use in the conversation with the user.

The sessionState field must be compressed using gzip and then base64 encoded before sending to Amazon Lex V2.

Request Body

The request accepts the following binary data.

inputStream

User input in PCM or Opus audio format or text format as described in the requestContentType parameter.

Response Syntax

HTTP/1.1 200 x-amz-lex-input-mode: inputMode Content-Type: contentType x-amz-lex-messages: messages x-amz-lex-interpretations: interpretations x-amz-lex-session-state: sessionState x-amz-lex-request-attributes: requestAttributes x-amz-lex-session-id: sessionId x-amz-lex-input-transcript: inputTranscript audioStream

Response Elements

If the action is successful, the service sends back an HTTP 200 response.

The response returns the following HTTP headers.

contentType

Content type as specified in the responseContentType in the request.

Length Constraints: Minimum length of 1.

inputMode

Indicates whether the input mode to the operation was text or speech.

Length Constraints: Minimum length of 1.

inputTranscript

The text used to process the request.

If the input was an audio stream, the inputTranscript field contains the text extracted from the audio stream. This is the text that is actually processed to recognize intents and slot values. You can use this information to determine if Amazon Lex V2 is correctly processing the audio that you send.

The inputTranscript field is compressed with gzip and then base64 encoded. Before you can use the contents of the field, you must decode and decompress the contents. See the example for a simple function to decode and decompress the contents.

Length Constraints: Minimum length of 1.

interpretations

A list of intents that Amazon Lex V2 determined might satisfy the user's utterance.

Each interpretation includes the intent, a score that indicates how confident Amazon Lex V2 is that the interpretation is the correct one, and an optional sentiment response that indicates the sentiment expressed in the utterance.

The interpretations field is compressed with gzip and then base64 encoded. Before you can use the contents of the field, you must decode and decompress the contents. See the example for a simple function to decode and decompress the contents.

Length Constraints: Minimum length of 1.

messages

A list of messages that were last sent to the user. The messages are ordered based on the order that you returned the messages from your Lambda function or the order that the messages are defined in the bot.

The messages field is compressed with gzip and then base64 encoded. Before you can use the contents of the field, you must decode and decompress the contents. See the example for a simple function to decode and decompress the contents.

Length Constraints: Minimum length of 1.

requestAttributes

The attributes sent in the request.

The requestAttributes field is compressed with gzip and then base64 encoded. Before you can use the contents of the field, you must decode and decompress the contents.

Length Constraints: Minimum length of 1.

sessionId

The identifier of the session in use.

Length Constraints: Minimum length of 2. Maximum length of 100.

Pattern: [0-9a-zA-Z._:-]+

sessionState

Represents the current state of the dialog between the user and the bot.

Use this to determine the progress of the conversation and what the next action might be.

The sessionState field is compressed with gzip and then base64 encoded. Before you can use the contents of the field, you must decode and decompress the contents. See the example for a simple function to decode and decompress the contents.

Length Constraints: Minimum length of 1.

The response returns the following as the HTTP body.

audioStream

The prompt or statement to send to the user. This is based on the bot configuration and context. For example, if Amazon Lex V2 did not understand the user intent, it sends the clarificationPrompt configured for the bot. If the intent requires confirmation before taking the fulfillment action, it sends the confirmationPrompt. Another example: Suppose that the Lambda function successfully fulfilled the intent, and sent a message to convey to the user. Then Amazon Lex V2 sends that message in the response.

Errors

For information about the errors that are common to all actions, see Common Errors.

AccessDeniedException

HTTP Status Code: 403

BadGatewayException

HTTP Status Code: 502

ConflictException

HTTP Status Code: 409

DependencyFailedException

HTTP Status Code: 424

InternalServerException

HTTP Status Code: 500

ResourceNotFoundException

HTTP Status Code: 404

ThrottlingException

HTTP Status Code: 429

ValidationException

HTTP Status Code: 400

Examples

Encode and decode a field

The following example provides two functions, one to compress a string with gzip and then base64 encode it and one to decode and decompress a string.

package com.amazonaws.deepsense.util; import com.amazonaws.deepsense.runtime.conversation.serialize.RuntimeSdkSerializer; import com.fasterxml.jackson.annotation.JsonInclude; import com.fasterxml.jackson.core.JsonProcessingException; import com.fasterxml.jackson.core.type.TypeReference; import com.fasterxml.jackson.databind.DeserializationFeature; import com.fasterxml.jackson.databind.ObjectMapper; import com.fasterxml.jackson.databind.SerializationFeature; import com.google.common.base.Preconditions; import java.io.IOException; import java.util.Base64; public class CompressionAndEncoding { private CompressionAndEncoding() { } /** * Given a generic object * 1. Serialize the object using jackson * 2. Compress the serialized object * 3. Base64 encode the compressed data */ public static String compressAndEncodeBase64(Object object) { if (object == null) { return null; } String objectAsString = GsonSerializer.serialize(object); byte[] compressedString = Compressions.compress(objectAsString); return Base64.getEncoder().encodeToString(compressedString); } /** * Given a base64 encoded, compressed, serialized object * 1. Base64 decodes the data * 2. Decompresses the base64 decoded data * 3. Converts the serialized object into a proper POJO */ public static <T> T decodeBase64AndDecompress(String objectAsString, TypeReference<T> typeRef) { if (objectAsString == null) { return null; } Preconditions.checkNotNull(typeRef, "Serialization class can't be null."); byte[] decodedBytes = Base64.getDecoder().decode(objectAsString); String decompressedObjectAsString = Compressions.uncompress(decodedBytes); try { return RuntimeSdkSerializer.instance().readValue(decompressedObjectAsString, typeRef); } catch (IOException e) { throw new RuntimeException(String.format("Unable to deserialize string %s", decompressedObjectAsString), e); } } }

See Also

For more information about using this API in one of the language-specific AWS SDKs, see the following: