基于百度AI接口的商品评论情感倾向性分析

开篇说明

阅读本篇文章前建议先将百度AI接口应用创建好，调试清楚，该流程在另一篇文章中介绍

百度AI接口应用创建与调试流程-CSDN博客

项目介绍

本项目接入两个百度AI接口（情感倾向性分析与机器翻译）对商品评论进行分析，得到结果后抽出置信度与积极消极概率再根据公式判断情感极性

项目架构

架构抽象为接口封装，只需要再Analysis.py中修改相关参数即可实现分析

规定Excel文档名称格式为：

sheet_name = "AmazonInfo_" + country + "_" + kind + ".xlsx"

项目代码

包括数据集一同打包在百度网盘，在pycharm中创建好新项目导入即可

链接：https://pan.baidu.com/s/1j4fvWpNUx1rse0VyDwAxNg?pwd=q7k7 
提取码：q7k7

文字介绍如下

Analysis.py

import json
import ApiConnect
import GetData
import Judge_confidence

# 定义分析哪一个Excel数据
# en_从英文翻译为中文,fr_从法语翻译为中文等等
lan_code = "en"
# 分析不同国家的亚马逊平台
country = "us"
# 分析亚马逊平台的商品
kind = "furniture"
# 数据总量
row_length = GetData.text(1, country, lan_code, kind)[1]
# 记录情感极性数量
mid_num = 0
neg_num = 0
pos_num = 0

for i in range(2, row_length):
    # 将判断数据转化为字典后获取其中包含了情感极性与置信度数据的items
    items = json.loads(ApiConnect.method(i, country, lan_code, kind)).get("items")
    # print(GetData.text(i, "us", "en", "furniture")[0])
    # print(items)
    # 获取情感极性几率与置信度
    conf = items[0].get("confidence")
    neg = items[0].get("negative_prob")
    pos = items[0].get("positive_prob")
    # 根据数据判断情感极性
    sentiment = Judge_confidence.determine_sentiment(conf, neg, pos)
    # print(sentiment)
    if sentiment == "积极":
        pos_num += 1
    elif sentiment == "消极":
        neg_num += 1
    else:
        mid_num += 1
    print("pos_num:", pos_num, "neg_num:", neg_num, "mid_num:", mid_num)

ApiConnect.py

这个代码应从百度AI接口的API文档中直接复制后再修改逻辑，由于我更改了他的方法调用参数，所以放出来以供参考

import requests
import json
import GetData

API_KEY = ""
SECRET_KEY = ""


def method(num_method, country, lan_code, kind):
    url = ("https://aip.baidubce.com/rpc/2.0/nlp/v1/sentiment_classify?charset=KuQwQKes8e2eM8EwmTiqCyMYO6yxSiWO"
           "&access_token=") + get_access_token()
    text = GetData.text(num_method, country, lan_code, kind)[0]

    payload = json.dumps(text)
    headers = {
        'Content-Type': 'application/json',
        'Accept': 'application/json'
    }

    response = requests.request("POST", url, headers=headers, data=payload)

    return response.text


def get_access_token():
    """
        使用 AK，SK 生成鉴权签名（Access Token）
        :return: access_token，或是None(如果错误)
    """
    url = "https://aip.baidubce.com/oauth/2.0/token"
    params = {"grant_type": "client_credentials", "client_id": API_KEY, "client_secret": SECRET_KEY}
    return str(requests.post(url, params=params).json().get("access_token"))

GetData.py

import json
import openpyxl
import LanTrans


# num_text 由Analysis定义，从第二行数据自动开始
def text(num_text, country, lan_code, kind):
    sheet_name = "AmazonInfo_" + country + "_" + kind + ".xlsx"
    wb = openpyxl.load_workbook(sheet_name)
    # 表名注意大小写
    sheet = wb["Sheet1"]
    row = sheet.max_row
    # 此处需要修改column，按照自己的excel文档确认，我这里是第六列
    cell_value = LanTrans.trans(sheet.cell(row=num_text, column=6).value, lan_code)
    cell_value_json = json.loads(cell_value)
    case_dict = {"text": cell_value_json["result"]["trans_result"][0]["dst"]}
    return case_dict, row

Judge_confidence.py

def determine_sentiment(confidence, negative_prob, positive_prob):
    if confidence > 0.7:
        if positive_prob > negative_prob:
            return "积极"
        else:
            return "消极"
    elif 0.4 < confidence <= 0.7:
        if positive_prob > 0.6:
            return "积极"
        elif negative_prob > 0.6:
            return "消极"
        else:
            return "中立/不明显"
    else:
        return "中立/不明显"

LanTrans.py

该接口代码修改逻辑与前面相同

import requests
import json

API_KEY = ""
SECRET_KEY = ""


def trans(value, lan_code):
    url = "https://aip.baidubce.com/rpc/2.0/mt/texttrans/v1?access_token=" + get_access_token()

    q = {"from": lan_code,
         "to": "zh",
         "q": value
         }
    payload = json.dumps(q)
    headers = {
        'Content-Type': 'application/json',
        'Accept': 'application/json'
    }

    response = requests.request("POST", url, headers=headers, data=payload)

    return response.text


def get_access_token():
    """
        使用 AK，SK 生成鉴权签名（Access Token）
        :return: access_token，或是None(如果错误)
    """
    url = "https://aip.baidubce.com/oauth/2.0/token"
    params = {"grant_type": "client_credentials", "client_id": API_KEY, "client_secret": SECRET_KEY}
    return str(requests.post(url, params=params).json().get("access_token"))

项目维护

预计更新：

输出情感极性判断结果到Excel中，含confidence，positive_prob，negative_prob，emo_result