2026-10-12T13:41:00Z https://www.scidb.cn/oai

10.57760/sciencedb.45942 2026-09-24T09:42:50Z

2026-09-24 Douban Detective Fiction Readers' Excerpt Text Data doi:10.57760/sciencedb.45942 en 1. Research Background and Data Overview1.1 Research BackgroundReading immersion is a core experience of deep reader–text interaction and a key phenomenon within narrative transportation theory. However, traditional studies have largely relied on reader self-reports or book reviews, which cannot easily capture readers' real-time text selection behavior during the reading process. This study treats reader-selected excerpts from the Douban book platform as "digital traces," employing computational text analysis methods to investigate which narrative content in detective fiction elicits heightened reader attention and immersive experience.1.2 Data SourceThis dataset originates from the Douban Books platform (book.douban.com). Douban is the largest book community in China, where users can highlight, excerpt, and share passages they find important or moving during the reading process. These excerpts result from readers' autonomous choices and reflect their attention to and emotional engagement with specific narrative content, thus serving as "digital traces" of reading immersion.1.3 Data Collection ScopeBook scope: The 100 highest-rated detective fiction novels on Douban.Time span: February 2011 to January 2024.Total excerpts: 27,999 reader-selected excerpts.Collection method: Publicly available reader excerpts were collected via Douban's public web scraping.2. Data File Description2.1 File Naming ConventionThis dataset contains one data file, which is the raw excerpt text dataset containing reader-selected excerpts and basic metadata.2.2 File FormatFormat: Microsoft Excel (.xlsx)Encoding: UTF-8Number of worksheets: One active worksheet (Sheet1)2.3 Column Name DefinitionsColumn 1: excerpted_text. Data type: Text (String). Description: Original passages excerpted by readers from detective fiction novels, autonomously selected and posted to the Douban platform.Column 2: time. Data type: Text (String). Description: The specific date and time when the excerpt was posted, in the format YYYY-MM-DD HH:MM:SS.3. Data Characteristics and DistributionThe data span a 14-year period from 2011 to 2024, with the period from 2020 to 2023 representing the peak data volume, accounting for approximately 70% of the total. The data deadline is around Jan 2024.4. Data Value and Reusability4.1 Data ValueMethodological innovation: This dataset treats reader-selected excerpts as "digital traces," offering an approach to measuring reading behavior that goes beyond traditional questionnaires and book reviews. The excerpts reflect "textual salience" during the reading process and serve as direct behavioral indicators of reading immersion.Multi-dimensional analytical potential: The dataset supports the combined use of multiple computational text analysis methods, including: topic modeling (LDA) to identify the thematic structure of narrative content; sentiment analysis to quantify the emotional orientation of texts; cognitive-semantic analysis to extract cognitive processing features of texts; and time-series analysis to explore diachronic changes in reader attention.Cross-disciplinary value: The dataset is applicable to multiple fields, including digital humanities, computational literary studies, reading psychology, and information behavior research.4.2 Data LimitationsThe data cover only the 100 highest-rated detective fiction novels on Douban and cannot represent all detective fiction or all types of literary works.The Douban user base exhibits demographic bias (predominantly young, highly educated urban residents), and excerpting behavior may be influenced by the platform's community culture.Excerpting behavior itself may be subject to social desirability effects, whereby users tend to select passages they consider "worth sharing" rather than those that genuinely moved them.4.3 Usage RecommendationsFor text classification or topic modeling, it is recommended to perform Chinese word segmentation and stop-word removal as preprocessing steps.The time field can be used for time-series analysis to explore diachronic trends in reader attention.Note: (Update) The examples have been withdrawn. If you intend to use this dataset for legitimate purposes such as learning or academic research, please contact the author to discuss your request. When applying for access, you are required to provide a data protection statement along with your relevant safeguards, specifying how the data will be stored, used, anonymized, and disposed of, and to commit that the data will only be used within the authorized scope and in a compliant manner. Access to the anonymised dataset will be granted only after the author has reviewed and reached an agreement with you. Douban; Excerp; Digital Reading Trace Tang Jiajun RESTRICTED dataset http://www.doi.org/10.1093/llc/fqag123 Science Data Bank