【python】SemEval 2014数据集 xml文件格式转换为csv+txt

笔记为自我总结整理的学习笔记，若有错误欢迎指出哟~

SemEval2014简介

SemEval2014，ABSA（ Aspect Based Sentiment Analysis）任务关注的领域是NLP中的细粒度情感分析，即给定一个句子判断其中的aspect以及它的情感极性。数据是基于laptop评论和restaurant评论

4个子任务

1.Aspect term extraction，方面术语抽取
2.Aspect term polarity，方面术语极性分类
3.Aspect category detection，方面类别抽取
4.Aspect category polarity，方面类别极性分类
对于子任务1和子任务2提供了laptop和restaurant数据，而对于子任务3和子任务4只提供了restaurant数据。

数据格式

原始数据为xml格式

xml转csv

xml_csv.py

import xml.etree.cElementTree as ET
import pandas as pddef xml_csv(listlist):xml = ['SemEval2014_Restaurants.xml']csv_name = ['SemEval2014_Restaurants.csv']# 解析XML文件tree = ET.parse(xml[listlist])root = tree.getroot()# 提取所有sentence元素sentences = root.findall('sentence')# 修复提取数据的方法，处理没有<aspectCategories>子元素的情况data = []# 遍历每个sentence元素for sentence in sentences:# 提取text内容text = sentence.find('text').text# 检查是否存在<aspectCategories>子元素aspect_categories_element = sentence.find('aspectCategories')if aspect_categories_element is not None:# 提取aspectCategories中的所有aspectCategory元素aspect_categories = aspect_categories_element.findall('aspectCategory')# 提取每个aspectCategories的term和polarityfor aspect_category in aspect_categories:category = aspect_category.get('category')polarity = aspect_category.get('polarity')data.append([text, category, polarity])df = pd.DataFrame(data, columns=['text', 'category', 'polarity'])df = df[df['polarity'].isin(['positive', 'negative', 'neutral'])]df['polarity'] = df['polarity'].map({'positive': 1, 'neutral': 0, 'negative': -1})df.to_csv(path_or_buf=csv_name[listlist], index=0)for i in range(1):xml_csv(i)

SemEval2014_Restaurants.csv

xml转txt

xml_txt.py

import xml.etree.cElementTree as ET# 写入文件
output_file = open('SemEval2014_Restaurants.txt', 'a', encoding='utf-8')def xml_txt(listlist):xml = ['SemEval2014_Restaurants.xml']csv_name = ['SemEval2014_Restaurants.csv']# 解析XML文件tree = ET.parse(xml[listlist])root = tree.getroot()# 提取所有sentence元素sentences = root.findall('sentence')# 修复提取数据的方法，处理没有<aspectCategories>子元素的情况data = []# 遍历每个sentence元素for sentence in sentences:# 提取text内容text = sentence.find('text').text# 检查是否存在<aspectCategories>子元素aspect_categories_element = sentence.find('aspectCategories')if aspect_categories_element is not None:# 提取aspectCategories中的所有aspectCategory元素aspect_categories = aspect_categories_element.findall('aspectCategory')# 提取每个aspectCategories的term和polarityfor aspect_category in aspect_categories:category = aspect_category.get('category')polarity = aspect_category.get('polarity')if polarity == "negative":polarity = -1elif polarity =="positive":polarity = 1else:polarity = 0# print(polarity)output_file.write(f"{polarity}\t{category}\t{text}\n")for i in range(1):xml_txt(i)

SemEval2014_Restaurants.txt