Sentiment analysis of sentences in python - Baidu Intelligent Cloud API - language processing technology

Keywords: Python api


Emotional analysis is used in my sister's graduation thesis. He has used crawler software to crawl various microblog accounts, and the text and comments on a topic are stored in excel.
Intends to use Baidu intelligent cloud, Natural Language Processing - > sentiment analysis.


Get access_token

API official documentation

To use these two parameters

  • client_id: required parameter, API Key applied;
  • client_secret: required parameter, the applied Secret Key;

The specific location of these two parameters:

  • First log in to the official website with baidu account:
  • Find your product: natural language processing
  • In the application, create a new application and enter a name and description

  • You can see that we need to copy the two parameters API Key and Secret Key, and replace the following code.
def get_token():
    API_Key = 'Copy and paste as above'
    Scret_Key = 'Copy and paste as above'
    url3 = ''
    data3 = {
        "grant_type":"client_credentials",     # Fixed value
        "client_id":API_KEY2,            #
    resp3 =,data=data3)
    print("access_token:" ,resp3.json()['access_token'])

Request emotion analysis interface

import requests
import json,time
import re,os

# Remove the HTML code and emoticons in the text through the re module 
def re_delete(content):
    dr = re.compile(r'<[^>]+>', re.S)
    emoji = re.compile("["u"\U0001F600-\U0001F64F"u"\U0001F300-\U0001F5FF"
                               u""u"\U0001F680-\U0001F6FF"u"\U0001F1E0-\U0001F1FF""]+", flags=re.UNICODE)
    content = dr.sub('', content)
    content = emoji.sub('', content)
    return content

# Request Baidu interface, return positive probability, confidence, negative probability, classification result 0 negative, 1 neutral, 2 positive
def fenxi(tex = "I love my country"):
    # headers = {'Content-Type': 'application/json'}        # You can ask for it without adding it
    tex = re_delete(tex)									# If your text does not have HTML or emoticons, you can not use this
    if not tex:
        return []
    access_token = '24.37f133deb9fefa877cf39583244079f8.2592000.1640158969.282335-25209811'
    url = f'{access_token}&charset=UTF-8'
    if len(tex.encode()) < 2048:							# The length of text bytes written in the document is up to 2048 bytes, which is almost 680 Chinese characters
        body = {'text' : tex}
        # Convert python dictionary type to json type
        body = json.dumps(body)
        # print(body)
        resp1 =,data=body)
            items = resp1.json()['items'][0]
            # round(x,3) means to keep the floating point number x to three decimal places
            return [round(items['positive_prob'],3), round(items['confidence'],3),
                    round(items['negative_prob'],3), items['sentiment']]
            print("The request is incorrect\n",resp1.text)
            # The printed error code can be compared with the official document
            return ["The request is incorrect"]
        print("Byte encoding length exceeds 2048\n",tex)
        # Because my text is generally not so long, I won't continue to analyze it. You can split the text here and call the interface
        return ["The length is too long"]
if __name__ == '__main__':
    # get_token()
    result = fenxi("I love my country")

Official documents: -p64nCQphmTY&Api=GWAI-7WcMrFnWb8M
You can view the parameter description and online call attempt at the above website, which is more detailed than that written in the previous document.

Read text operation

  • The request interface is successful, and then the difference code reads the text in batch to make a request
  • Everyone's text is different. Maybe csv,excel,txt and so on will not be written in detail here
  • A reminder is that Baidu has a QPS limit, that is, it can request several times a second at most. If there are many requests, it will return an error
  • When you request, just add time.sleep(0.5). The number of 0.5 is 1/qQPS
  • Of course, you can also spend money. There will be a lot of requests. You can find more Baidu accounts and get more access_ Just a token
  • pycharm can run multiple py codes at the same time (there are still restrictions on applying for access_token s of different applications for the same account)

  • Thank you for reading. If you think it's useful, you might as well give a compliment before you go

Posted by Dollar on Mon, 22 Nov 2021 15:51:53 -0800