How to Deploy and Integrate Azure Cognitive Services: Computer Vision and Text Analytics
Experiment #102 – Discover How to Deploy and Integrate Azure Cognitive Services: Computer Vision and Text Analytics
1.
Create the Computer Vision Resource
Computer Vision is an Azure Cognitive Service that allow us to get information from an image. It analyzes the image environment, detects objects inside them and classifies all of them by assigning tags and labels. We can find this resource in our Azure Portal under the “AI + Machine Learning” group:
After deploying the service, we must look and write down at the subscription key and the endpoint to integrate the service in our project.
2. Create the Text Analytics Resource
Text Analytics is an Azure Cognitive Service that allow us to get information from a text. It analyzes the text context, detects subjects and topics and classifies all of them by assigning tags and labels. We can also find this resource in our Azure Portal under the “AI + Machine Learning” group:
After deploying the service, we must look and write down at the subscription key and the endpoint to integrate the service in our project.
3. Get caption and tags from Images
Now, we are going to integrate the HTTP requests to the Computer Vision API to send it an image and get caption and tags:
subscription_key = 'xxxxxxxxx'
assert subscription_key
vision_endpoint = 'https://xxxxxxxxx.cognitiveservices.azure.com/'
vision_base_url = vision_endpoint + "vision/v1.0/"
vision_analyze_url = vision_base_url + "analyze"
images_folder = "https://raw.githubusercontent.com/idiWork/Experiment_102/master/images/idiwork_experiment_102_sample_image_"
resort_pool = images_folder + "01.jpg"
resort_room = images_folder + "02.jpg"
resort_diner = images_folder + "03.jpg"
hotel_diner = images_folder + "04.jpg"
hotel_pool = images_folder + "05.jpg"
hotel_room = images_folder + "06.jpg"
image_url = resort_room
from IPython.display import Image, display
display(Image(image_url))
import requests
headers = {'Ocp-Apim-Subscription-Key': subscription_key }
params = {'visualFeatures': 'Categories,Description,Tags,Color'}
data = {'url': image_url}
response = requests.post(vision_analyze_url, headers=headers, params=params, json=data)
response.raise_for_status()
analysis = response.json()
analysis
caption = analysis["description"]["captions"][0]["text"].capitalize()
caption
topTags = analysis["description"]["tags"][0:4]
topTags
4. Get sentiment from Text
Then, we are going to implement the HTTP requests to the Text Analytics API to send it a text and get the overall sentiment:
text_analytics_subscription_key = 'xxxxxxxxx'
assert text_analytics_subscription_key
text_analytics_base_url = 'https://xxxxxxxxx.cognitiveservices.azure.com/text/analytics/v2.0/'
sentiment_api_url = text_analytics_base_url + "sentiment"
neg_sent = """
This is the worst stay I have every had! I have nothing positive to say about the diner and am extremely angry.
Poor service, bad food and too expensive drinks. I will not recommend it to anyone.
"""
pos_sent = """
I would like to thank you very much for a wonderful stay at the hotel.
The room we stayed in was very nice and had plenty of room for the whole family and the beds were especially comfortable.
"""
neutral_sent = """
The kids go to the swimming pool all day while we take sunbaths near them.
We have an umbrella and a deck chair for each of us if we go early in the morning.
"""
long_review = """
We arrived in the evening and the reception was very busy,
so, we decided to wait until it was quieter.
Unfortunately, we could not charge anything to the room yet as we were not checked in.
We could not use the Wi-Fi either so had to use the guest Wi-Fi which was very slow and poor quality.
Once we checked in, the room was very nice and had fantastic views, but the bathroom was quite dirty.
We called reception and they sent housekeeping who were very quick to clean the bathroom properly and even gave us some free bath toys for the kids which they loved.
Overall the stay was pleasant and the service from staff was excellent.
"""
review_text = long_review
documents = {'documents' : [
{'id': '1', 'language': 'en', 'text': review_text}
]}
headers = {"Ocp-Apim-Subscription-Key": text_analytics_subscription_key}
response = requests.post(sentiment_api_url, headers=headers, json=documents)
sentiments = response.json()
sentiments
score = sentiments['documents'][0]['score']
score
score_interpretation = "Neutral"
if (score < 0.4):
score_interpretation = "Negative"
elif (score >= 0.6):
score_interpretation = "Positive"
score_interpretation
5. Invoke the Deployed Services
In this step we are going to invoke the services previously created and deployed to use them in combination with the Azure Cognitive Services:
def invoke_service(ml_service_key, ml_service_scoring_endpoint, ml_service_input):
headers = {"Authorization": "Bearer " + ml_service_key}
response = requests.post(ml_service_scoring_endpoint, headers=headers, json=ml_service_input)
result = response.json()
return result
classifier_service_key = ""
classifier_service_scoring_endpoint = 'http://xxx-xxx-xxx-xxx-xxx.westeurope.azurecontainer.io/score'
classifier_service_input = [review_text]
classifier_result = invoke_service(classifier_service_key, classifier_service_scoring_endpoint, classifier_service_input)
classifier_result
classification = 'Room Review'
if classifier_result == 1:
classification = 'Diner Review'
elif classifier_result == 2:
classification = 'Pool Review'
classification
summarizer_service_key = ""
summarizer_service_scoring_endpoint = 'http://xxx-xxx-xxx-xxx-xxx.westeurope.azurecontainer.io/score'
summarizer_service_input = review_text
summarizer_result = invoke_service(summarizer_service_key, summarizer_service_scoring_endpoint, summarizer_service_input)
formatted_result = summarizer_result[0].replace("\\n", " ").strip() if len(summarizer_result) > 0 else "N/A"
formatted_result
6. Summarize the Results
Finally, we will create and apply an HTML template to display the results:
from IPython.core.display import HTML
displayTemplate = """
<div><h3>Claim Summary</h3></div>
<div> </div>
<div><b>Classification:</b> {}</div>
<div><b>Caption:</b> {}</div>
<div><b>Tags:</b> {}</div>
<div><b>Sentiment:</b> {}</div>
<div> </div>
<div><img src='{}' width='200px'></div>
<div> </div>
<div><b>Summary:</b></div>
<div>{}</div>
<div> </div>
<div><b>Claim:</b></div>
<div>{}</div>
"""
displayTemplate = displayTemplate.format(classification, caption, ' '.join(topTags), score_interpretation, image_url, formatted_result, review_text)
display(HTML(displayTemplate))
In the following animation you will see the processes and outputs of running these steps in Azure Notebooks with Jupyter:
