作者：MeshCloud脉时云公有云架构师陈博文

简介：

Dialogflow 是Google 提供的一款人机交互平台，通过该平台可以轻松地设计出属于自己的交互机器人，比如常见的网页聊天机器人，电话智能客服等。借助Dialogflow甚至可以用于扫地机器人交互系统或者更高级的使用。

Dialogflow 通过客户输入的语音或者文字甚至情感分析，来识别客户的意图（Intens)，结合实体(Entities)，来进行相应的回复。

Dialogflow的几个优点：

• 识别准确率高，响应速度快
• 支持 30 多种语言和语言变体
• 上手简单：图形界面配置；官方文档丰富、详细；网络上有案例可供参考
• 有问题易解决：开发者社区超过150万名开发者

Dialogflow经典案例：

一、马航的订票查票机器人：

使用 Google Cloud 上的 Dialogflow，马来西亚航空公司和 Amadeus 创建了一个聊天机器人，使客户能够搜索、预订和支付航班，从而使航空公司能够满足未来的需求并增加数字渠道的收入。

二、达美乐披萨的订餐机器人：

三、KLM预定、打包机器人：

KLM 于 2016 年开始探索为客户提供体验的方法。他们在测试多个平台后选择了 Dialogflow。

常用工具

一、内置 Small Talk

Small Talk 用于为闲聊对话提供响应。 此功能可以解答代理范围之外的常见问题，极大地提升最终用户体验。

Small Talk 有两种版本：

• 内置 Small Talk：为代理启用 Small Talk 后，它会自动处理闲聊对话，无需向代理添加意图。

• 预建 Small Talk：导入预建 Small Talk 代理时，它会提供处理闲聊对话的意图。

二、prebuilt agent

由 Dialogflow 提供的一组代理，适用于常见的使用场景。 您可以这些代理为基础，构建涵盖特定场景（如外出就餐、酒店预订和导航）的对话。

由 Dialogflow 提供的一组代理，适用于常见的使用场景。 您可以这些代理为基础，构建涵盖特定场景（如外出就餐、酒店预订和导航）的对话。

如何制作一个自己的天气&新闻语音问答机器人

使用了文字输入Dialogflow 的方式

通过speech-to-text将音频麦克风流到Dialogflow 的文本意图检测API

案例使用了以下GCP产品：

• Dialogflow ES & Knowledge Bases

• Speech to Text

其它组件：

• Webhook

• Weathers & News API

在这个demo中你可以使用麦克风输入，然后返回新闻或者天气

一、Dialogflow ES（页面配置）

1、意图配置

①配置输入

②配置回复

2、Webhook配置

①意图开启Fulfillment

②添加webhook

③webhook代码

import requests
#新闻接口
from newsapi import NewsApiClient
import time
import json
#使用了flask框架
from flask import Flask, request
import pycountry
#from gevent.pywsgi import WSGIServer

app = Flask(__name__)

@app.route('/webhook', methods=['POST'])
def webhook():
    Dialogflow_data = json.loads(request.data)
    intent =Dialogflow_data["queryResult"]["intent"]["displayName"]
    print("--------------------------------------")
    if intent == "news":
        responseText = callnewsapi()
        news = responseText["articles"][0]["title"]
        print(news)
        headline = " headline news is %s"%(news)
        #需要按要求返回dialogflow才能回复给客户端
        #"fulfillmentText"是客户端接收消息
        res = {"fulfillmentText": headline ,"fulfillmentMessages": [{"text": {"text":[headline]}}]}
        return(res)
    elif intent == "weather":
        CITY=Dialogflow_data["queryResult"]["parameters"]["geo-city"]
        key = '479284d0d8574437b8170935221508'
        responseText = json.loads(callweatherapi(key,CITY))
        mintemp = responseText["data"]["ClimateAverages"][0]["month"][7]["avgMinTemp"]
        maxtemp = responseText["data"]["ClimateAverages"][0]["month"][7]["absMaxTemp"]
        tempres = "London Maxtemp is %s ℃ Mintempe is %s ℃"%(maxtemp,mintemp)
        #需要按要求返回dialogflow才能回复给客户端
        #"fulfillmentText"是客户端接收消息
        res = {"fulfillmentText": tempres ,"fulfillmentMessages": [{"text": {"text":[tempres]}}]}
        return(res)


def callweatherapi(key,CITY):
    time.sleep(0.01)
    response = requests.post("http://api.worldweatheronline.com/premium/v1/weather.ashx?key=%s&q=%s&fx=no&cc=no&mca=yes&format=json"%(key,CITY))
    if response.status_code == 200:
       return(response.text)

def callnewsapi():
    newsapi = NewsApiClient(api_key='0eaad3923a654da2a2a32d84870e0405')
    response = newsapi.get_top_headlines(language='es')
    return(response)

if __name__ == '__main__':
    #WSGIServer(('0.0.0.0', 5000), app).serve_forever()
    app.run(host="0.0.0.0", port=5000, ssl_context=('/root/scs1660552637313_cbw404.cn/scs1660552637313_cbw404.cn_Nginx/scs1660552637313_cbw404.cn_server.crt', '/root/scs1660552637313_cbw404.cn/scs1660552637313_cbw404.cn_Nginx/scs1660552637313_cbw404.cn_server.key'))

新闻接口:

http://api.worldweatheronline.com/premium/v1/weather.ashx?key=apikey&q=city&fx=no&cc=no&mca=yes&format=json

天气接口:

#install
pip install newsapi-python
#usage
from newsapi import NewsApiClient
#init
newsapi = NewsApiClient(api_key='API_KEY')\
# /v2/top-headlines
top_headlines = newsapi.get_top_headlines(q='bitcoin',
                                          sources='bbc-news,the-verge',
                                          category='business',
                                          language='en',
                                          country='us')
# /v2/everything
all_articles = newsapi.get_everything(q='bitcoin',
                                      sources='bbc-news,the-verge',
                                      domains='bbc.co.uk,techcrunch.com',
                                      from_param='2017-12-01',
                                      to='2017-12-12',
                                      language='en',
                                      sort_by='relevancy',
                                      page=2)
# /v2/top-headlines/sources
sources = newsapi.get_sources() 

二、Speech-to-text(后面简称stt) to Dialogflow

1、准备工作

①权限配置

下载service account json格式

Linux:
配置环境变量 export GOOGLE_APPLICATION_CREDENTIALS=
 为1中下载的 sa 的json文件

Windows:
set GOOGLE_APPLICATION_CREDENTIALS=C:\Users\Administrator\Downloads\sa.json

②python包

python 包
google-cloud-speech
pyaudio
google-cloud-dialogflow
python-dotenv
uuid

③.env文件 用于读取配置

PROJECT_ID=
#这里做的西班牙语测试
LANGUAGE_CODE=es
#语音的一些参数设置，保持默认
ENCODING=AUDIO_ENCODING_LINEAR_16
SAMPLE_RATE_HERZ=16000
SINGLE_UTTERANCE=false
SPEECH_ENCODING=LINEAR16
SSML_GENDER=FEMALE
#dialogflow的区域（有us，es，zh)
LOCATION_ID=global

2、Speech-to-text

使用实时流式音频执行识别（transcribe_streaming_mic）,也就是麦克风持续输入,代码如下：

#!/usr/bin/env python

"""Google Cloud Speech API sample application using the streaming API.
NOTE: This module requires the additional dependency `pyaudio`. To install
using pip:
    pip install pyaudio
Example usage:
    python transcribe_streaming_mic.py
"""

from __future__ import division

import re
import sys

from google.cloud import speech

import pyaudio
from six.moves import queue


RATE = 16000
CHUNK = int(RATE / 10)  # 100ms


class MicrophoneStream(object):
    """Opens a recording stream as a generator yielding the audio chunks."""

    def __init__(self, rate, chunk):
        self._rate = rate
        self._chunk = chunk


        self._buff = queue.Queue()
        self.closed = True

    def __enter__(self):
        self._audio_interface = pyaudio.PyAudio()
        self._audio_stream = self._audio_interface.open(
            format=pyaudio.paInt16,

            channels=1,
            rate=self._rate,
            input=True,
            frames_per_buffer=self._chunk,
            stream_callback=self._fill_buffer,
        )

        self.closed = False

        return self

    def __exit__(self, type, value, traceback):
        self._audio_stream.stop_stream()
        self._audio_stream.close()
        self.closed = True
        self._buff.put(None)
        self._audio_interface.terminate()

    def _fill_buffer(self, in_data, frame_count, time_info, status_flags):
        """Continuously collect data from the audio stream, into the buffer."""
        self._buff.put(in_data)
        return None, pyaudio.paContinue

    def generator(self):
        while not self.closed:
            chunk = self._buff.get()
            if chunk is None:
                return
            data = [chunk]
            while True:
                try:
                    chunk = self._buff.get(block=False)
                    if chunk is None:
                        return
                    data.append(chunk)
                except queue.Empty:
                    break

            yield b"".join(data)


def listen_print_loop(responses):
    num_chars_printed = 0
    for response in responses:
        if not response.results:
            continue

        result = response.results[0]
        if not result.alternatives:
            continue

        transcript = result.alternatives[0].transcript

        overwrite_chars = " " * (num_chars_printed - len(transcript))

        if not result.is_final:
            sys.stdout.write(transcript + overwrite_chars + "\r")
            sys.stdout.flush()

            num_chars_printed = len(transcript)

        else:
            print(transcript + overwrite_chars)

            if re.search(r"\b(exit|quit)\b", transcript, re.I):
                print("Exiting..")
                break

            num_chars_printed = 0


def main():

    language_code = "en-US"  #  BCP-47 

    client = speech.SpeechClient()
    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=RATE,
        language_code=language_code,
    )

    streaming_config = speech.StreamingRecognitionConfig(
        config=config, interim_results=True
    )

    with MicrophoneStream(RATE, CHUNK) as stream:
        audio_generator = stream.generator()
        requests = (
            speech.StreamingRecognizeRequest(audio_content=content)
            for content in audio_generator
        )

        responses = client.streaming_recognize(streaming_config, requests)

        # Now, put the transcription responses to use.
        listen_print_loop(responses)


if __name__ == "__main__":
    main()

3、Dialogflow

调用检测意图，代码如下:

#!/usr/bin/env python

"""DialogFlow API Detect Intent Python sample to use regional endpoint.
Examples:
  python detect_intent_texts_with_location.py -h
  python detect_intent_texts_with_location.py --project-id PROJECT_ID \
  --location-id LOCATION_ID --session-id SESSION_ID \
  "hello" "book a meeting room" "Mountain View"
"""

import argparse
import uuid



def detect_intent_texts_with_location(
    project_id, location_id, session_id, texts, language_code
):

    from google.cloud import dialogflow

    session_client = dialogflow.SessionsClient(
        client_options={"api_endpoint": f"{location_id}-dialogflow.googleapis.com"}
    )

    session = (
        f"projects/{project_id}/locations/{location_id}/agent/sessions/{session_id}"
    )
    print(f"Session path: {session}
")

    text_input = dialogflow.TextInput(text=texts, language_code=language_code)

    query_input = dialogflow.QueryInput(text=text_input)

    response = session_client.detect_intent(
        request={"session": session, "query_input": query_input}
    )

    print("=" * 20)
    print(f"Query text: {response.query_result.query_text}")
    print(
        f"Detected intent: {response.query_result.intent.display_name} (confidence: {response.query_result.intent_detection_confidence,})
"
    )
    print(f"Fulfillment text: {response.query_result.fulfillment_text}
")




if __name__ == "__main__":
    parser = argparse.ArgumentParser(
        description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter
    )
    parser.add_argument(
        "--project-id", help="Project/agent id.  Required.", required=True
    )
    parser.add_argument("--location-id", help="Location id.  Required.", required=True)
    parser.add_argument(
        "--session-id",
        help="Identifier of the DetectIntent session. " "Defaults to a random UUID.",
        default=str(uuid.uuid4()),
    )
    parser.add_argument(
        "--language-code",
        help='Language code of the query. Defaults to "en-US".',
        default="en-US",
    )
    parser.add_argument("texts", nargs="+", type=str, help="Text inputs.")

    args = parser.parse_args()

    detect_intent_texts_with_location(
        args.project_id,
        args.location_id,
        args.session_id,
        args.texts,
        args.language_code,
    )

4、（主要代码）将stt的结果（文字）输出到Dialogflow 意图检测，Dialogflow作出回复

流程：

代码如下：

#!/usr/bin/env python

"""Google Cloud Speech API sample application using the streaming API.
NOTE: This module requires the additional dependency `pyaudio`. To install
using pip:
    pip install pyaudio
Example usage:
    python transcribe_streaming_mic.py
"""

from __future__ import division

import re
import sys

from google.cloud import speech

import pyaudio
from six.moves import queue
import os
import uuid
#调用 Dialogflow意图检测包（代码见2.dialogflow）
from detect_intent_texts_with_location import detect_intent_texts_with_location
from dotenv import load_dotenv

RATE = 16000
CHUNK = int(RATE / 10)  # 100ms


class MicrophoneStream(object):
    """Opens a recording stream as a generator yielding the audio chunks."""

    def __init__(self, rate, chunk):
        self._rate = rate
        self._chunk = chunk

        
        self._buff = queue.Queue()
        self.closed = True

    def __enter__(self):
        self._audio_interface = pyaudio.PyAudio()
        self._audio_stream = self._audio_interface.open(
            format=pyaudio.paInt16,
            channels=1,
            rate=self._rate,
            input=True,
            frames_per_buffer=self._chunk,
            stream_callback=self._fill_buffer,
        )

        self.closed = False

        return self

    def __exit__(self, type, value, traceback):
        self._audio_stream.stop_stream()
        self._audio_stream.close()
        self.closed = True
        self._buff.put(None)
        self._audio_interface.terminate()

    def _fill_buffer(self, in_data, frame_count, time_info, status_flags):
        """Continuously collect data from the audio stream, into the buffer."""
        self._buff.put(in_data)
        return None, pyaudio.paContinue

    def generator(self):
        while not self.closed:
            chunk = self._buff.get()
            if chunk is None:
                return
            data = [chunk]
            while True:
                try:
                    chunk = self._buff.get(block=False)
                    if chunk is None:
                        return
                    data.append(chunk)
                except queue.Empty:
                    break

            yield b"".join(data)


def listen_print_loop(responses):

    load_dotenv(verbose=True)
    num_chars_printed = 0
    for response in responses:
        if not response.results:
            continue


        result = response.results[0]
        if not result.alternatives:
            continue

        transcript = result.alternatives[0].transcript

        overwrite_chars = " " * (num_chars_printed - len(transcript))

        if not result.is_final:
            sys.stdout.write(transcript + overwrite_chars + "\r")
            sys.stdout.flush()

            num_chars_printed = len(transcript)

        else:
        #从.env中导出Project_id等配置，可以通过修改.env修改
            TEXT=transcript + overwrite_chars
            print(transcript + overwrite_chars)
            PROJECT_ID = os.getenv("PROJECT_ID")
            SESSION_ID = uuid.uuid1()
            LANGUAGE_CODE = os.getenv("LANGUAGE_CODE")
        #Location_ID
            LOCATION_ID = os.getenv("LOCATION_ID")
        #意图检测 TEXT为mic接收到的语音转成的文字（代码见2.dialogflow）
            detect_intent_texts_with_location(PROJECT_ID, LOCATION_ID, SESSION_ID, TEXT, LANGUAGE_CODE)
            # Exit recognition if any of the transcribed phrases could be
            # one of our keywords.
            #对麦克风说exit即可退出
            if re.search(r"\b(exit|quit)\b", transcript, re.I):
                print("Exiting..")
                break

            num_chars_printed = 0


def main():
    language_code = "en-US"  # BCP-47 
    client = speech.SpeechClient()
    config = speech.RecognitionConfig(
        encoding=speech.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz=RATE,
        language_code=language_code,
    )

    streaming_config = speech.StreamingRecognitionConfig(
        config=config, interim_results=True
    )

    with MicrophoneStream(RATE, CHUNK) as stream:
        audio_generator = stream.generator()
        requests = (
            speech.StreamingRecognizeRequest(audio_content=content)
            for content in audio_generator
        )

        responses = client.streaming_recognize(streaming_config, requests)


        listen_print_loop(responses)

if __name__ == "__main__":
    main()

Location_id:（上面意图检测API的location_id参数）

国家/地区分组	地理位置	地区 ID
美洲	爱荷华	us-central1
美洲	蒙特利尔	northamerica-northeast1
美洲	南卡罗来纳	us-east1
美洲	俄勒冈	us-west1
欧洲	比利时	europe-west1
欧洲	伦敦	europe-west2
欧洲	法兰克福	europe-west3
亚太地区	悉尼	australia-southeast1
亚太地区	东京	asia-northeast1
亚太地区	孟买	asia-south1
亚太地区	新加坡	asia-southeast1
全球	全球服务，静态数据在美国	global（首选）、us 或无区域（默认）

5、测试

①Dialogflow web测试：Fulfillment 通过 webhook response 格式化数据后返回给客户端 ，测试成功

②程测试：mic -- stt -- dialogflow --client（default welcome intent --default response）

测试成功：

可以看到语音输入的内容转成了文字，并发送给了dialogflow的意图检测，并匹配意图给出相应的回复

③全流程测试：mic -- stt -- dialogflow -- fulfillment -- webhook -- api -- client

对麦克风说：noticias(西语新闻）

返回：头条新闻的标题，测试成功

三、总结

至此，一个天气&新闻语音问答机器人就做好了

官方还提供了另外的集成或者使用方式，可以供大家参考学习。希望这篇文章对大家有抛砖引玉的作用，以便大家能做出更高级更智能、更符合自己项目的交互机器人