Psycholinguistic LIWC and n-gram counts in a corpus of 1145 English and Dutch novels

This dataset consists of CSV files with word counts in several corpora: - 694 English language novels from male and female authors classified by authors' sexual orientation (heterosexual, bisexual, homosexual) - 401 bestselling Dutch language novels - 50 novels nominated for Dutch literary prizes Each corpus comes with: - LIWC counts; this file also includes the available metadata for each novel. The English data was created with LIWC 2015. The Dutch data was created with the validated translation of LIWC 2001. - Word counts (unigrams) and bigram counts per novel. All text has been converted to lowercase. Contractions are tokenized into separate tokens, e.g., can't => ca n't Two restrictions are applied: - only unigrams or bigrams that occur in at least 10 texts are retained - only the 100k most frequent are retained - Overall word counts and bigram counts; i.e., the sum across all novels. All files are encoded in UTF-8. The word counts were extracted with the countngrams.py script.

本数据集包含多个语料库的词频统计CSV文件：
- 694部英语小说，作者涵盖男女，且按作者性取向分为异性恋、双性恋、同性恋三类；
- 401部畅销荷兰语小说；
- 50部获荷兰文学奖提名的小说。

每个语料库均附带以下内容：
- 语言学调查与词数统计（Linguistic Inquiry and Word Count，简称LIWC）计数文件：该文件同时包含每部小说的可用元数据。其中英语数据采用LIWC 2015生成，荷兰语数据则基于经过验证的LIWC 2001译版生成。
- 每部小说的单字（unigrams）词频统计与二元组（bigrams）计数。所有文本均已转换为小写形式，缩写会被拆分为独立的Token（Token），例如can't将被拆分为ca n't。

共应用两项筛选规则：
1. 仅保留在至少10部文本中出现过的单字或二元组；
2. 仅保留频率排名前100,000的单元。

此外还包含整体词频与二元组计数，即所有小说的统计总和。所有文件均采用UTF-8编码，词频统计通过countngrams.py脚本提取生成。