结构主题模型（二）复现

简介：整理在复现论文（stm An R Package for Structural Topic Models）代码过程中的注意事项和报错信息

结构主题模型专题文章：

复现代码

下载代码及数据：

导包

stm包依赖R(≥ 3.5.0)

install.packages("stm")

导入数据

Warning messages: 1: In read.table(file = file, header = header, sep = sep, quote = quote,  :  invalid input found on input connection data_preprocessing_test1a.csv 
2: In read.table(file = file, header = header, sep = sep, quote = quote,  :  incomplete final line found by readTableHeader on data_preprocessing_test1a.csv

问题二：列名和列的数量不一致

Error in read.table(file = file, header = header, sep = sep, quote = quote, : more columns than column names

问题三：EOF with quoted string

In scan(file = file, what = what, sep = sep, quote = quote, dec = dec,  :   EOF within quoted string

问题四：对于中文文本，易因换行/n等其他分隔符使得程序对csv文件解析错误

针对以上问题，建议使用.xlsx或./xls格式的数据

1. 先使用python将csv转excel

import pandas as pd
csv_file = pd.read_csv(data.csv, low_memory=False, encoding=utf-8)
csv_file.to_excel(data.xlsx, index=False, encoding=utf-8)

2. R导入excel数据

install.packages("readxl")
library(readxl)
data <- read_excel(path="data.xlsx", sheet="your_sheet_name", col_names= TRUE)

代码执行

问题一

system is computationally singular: reciprocal condition number = 1.06304e-19

可能的原因：存在矩阵不可逆，对于矩阵A而言，det(A)=/≈0都会使得A不可逆。

R中函数det()用于计算矩阵的行列式

问题二

K=2 is equivalent to a unidimensional scaling model which you may prefer.

选择主题数时不宜选择等于或低于2的主题数，这样的话，结构主题模型就不能发挥较大的作用

问题三

文本数量和模型中的文档数量不匹配

Error in findThoughts(out.stm, topics = 27, texts = corpus$documents$texts, : Number of provided texts and number of documents modeled do not match

可能的解决方法（未解决博主的问题）：