ABSTRACT This paper presents four case studies guided by the research questions: how do doctoral students distribute their citation forms in the same literary practice, how do these distributions change across different versions of the text produced during a year and a half of doctoral training, and what reasons do they attribute to these choices and preferences of use? To answer them, we analyzed the style and integration of citations in three versions per writer, and contrasted with semi-structured interviews. This combined methodological approach contributes to the growing tradition of postgraduate writing studies interested in the relationships between the writer’s discursive choices and the construction of his or her academic identity.

The academic writing software market is booming, projected to reach [insert projected value for 2033] by 2033, driven by increasing online learning and the need for plagiarism detection tools. This comprehensive analysis explores market trends, key players (Grammarly, Turnitin, etc.), and regional growth, offering valuable insights for businesses and educators.

Description of Supporting Files Demography, education, and research trends in the interdisciplinary field of disease ecology Ellen E. Brandell, Daniel J. Becker, Laura Sampson, Kristian M. Forbes TopArticles_Inclusion.xlsx This Excel provides a list of influential articles written in by survey participants at least two times. Sheet “table”: just tabular information Sheet “withNotes”: includes notes about data, number of citations from survey participants, and percent inclusion calculations. Columns are: ‘INCLUDED’: if the article appeared in the corpus (1) or not (0) ‘COUNT’: the number of times survey participants wrote in the article ‘ARTICLE’: article citation Percent of articles included in the corpus are calculated for 4 or more write-ins, 3-write-ins, 2 write-ins, and across all articles written in twice. IRB_Correspondence_STUDY00010582.pdf Institutional Review Board correspondence and approval from Pennsylvania State University. Survey response data may be available upon request from the corresponding author. To protect participants, any potentially identifying information will be removed prior to filling a request. See the online Supporting Information for this article for extensive reporting of survey results prior to a request. FullSurvey.pdf A PDF of the full survey form. CorpusFrequencyAnalysis.ipynb This is the Python script used for corpus organization and the topic detection analysis. It includes some plot generation.

Description The dataset consists of 600 text-only prompts, each representing a fine-tuned instance of a single-turn user exchange in English. The samples are categorized into 10 distinct classes and cover 19 specific use cases. The dataset has been generated using ethically sourced human-in-the-loop data generation methods involving detailed insights of subject matter experts on labeled data for supervised fine-tuning to map input text with corresponding output responses.

The dataset is beneficial for direct preference optimization to generate responses that reinforce learning through human feedback. These techniques have been applied to align the fine-tuned conversational prompts with the desired output characteristics to ensure coherence, relevance, and alignment with the specified use cases and categories.

Key Features

User Intent-Centric Prompts: Prompts are designed primarily to capture user intent and are formulated using natural language processing techniques. Conversational Interactions: The dataset facilitates interactive dialogues addressing a diverse range of queries in areas such as writing assistance, coding support, knowledge retrieval, data manipulation, logical reasoning, and classification tasks. Dataset Source Subject matter expert annotators @SoftAgeAI have annotated the data at simple and complex levels, focusing on quality factors such as content accuracy, clarity, coherence, grammar, depth of information, and overall usefulness.

Structure & Fields The dataset is organized into five columns, which are detailed below:

S No (int64): A sequential identifier for each prompt, ranging from 1 to 600. Prompts (object): The text of the prompt or query, which is the input given by the user. These prompts cover a wide range of topics, including shopping assistance, creative writing, Q&A, and more. Use-cases (object): Describes the primary use case or application of the prompt. This categorization includes roles such as "Shopping assistant," "Creative writing assistant," "Q&A helper," and "Specialized knowledge helper." Type (object): Indicates the complexity or nature of the prompt, with all entries in this dataset labeled as "Simple." Categories (object): Provides a broader categorization of the prompt, such as "Open ended QA" or "Writing," offering additional context on the expected interaction or outcome. Intended Use Cases

The dataset is designed to improve the functionality of query assistance models across various domains, including coding, creative writing, travel support, marketing recommendations, citation management, academic writing, language translation, logical reasoning, research assistance, specialized knowledge-related, and STEM-related applications. The dataset aims to facilitate the development of generative models in fields such as e-commerce, customer support, educational applications, user query suggestions, and general-purpose chatbots. It is suitable for pre-training large language models utilizing supervised, fine-tuned annotated data and retrieval-augmented generative models. The dataset is curated to exclude interactions involving violence, harm, conflict, discrimination, brutality, and misinformation to ensure ethical use in its intended applications. Potential Limitations & Biases This is a static dataset, so the information is dated May 2024.

Note If you have any questions related to our data annotation and human review services for large language model training and fine-tuning, please contact us at SoftAge Information Technology Limited at info@softage.ai.

Bangladeshi Writer Identification Dataset

High-quality Bangladeshi handwriting for writer identification research

Short summary

This dataset contains 424 standardized RGB JPG handwriting images collected from 91 Bangladeshi university students. Images were cleaned and normalized for reproducible writer identification and document-image experiments. Some writers have fewer than five usable samples; unreadable or corrupted files were omitted during processing.

Dataset overview

• Writers: 89
• Total images: 424
• Average images per writer: ~4.66
• Image format: JPG (RGB)
• Max dimension: 1024 × 1024 (aspect-ratio preserved)
• JPEG quality: 95
• License: CC BY-NC-SA 4.0

Important note on image count

The raw collection originally targeted five samples per writer (91 × 5 = 455). During cleaning and standardization some source files were missing or unreadable; the released dataset therefore contains 424 verified, usable images. This dataset represents the cleaned, usable data rather than a forced completion of missing samples.

Contents and file layout

standardized_dataset/
├── images/        # standardized JPG images
├── metadata.csv     # per-image metadata (image_path, writer_id, writer_number, ...)
├── README.md       # dataset overview and usage notes
├── DESCRIPTION.md    # (this file) full Kaggle-ready description
├── dataset_info.json   # processing summary
└── LICENSE        # CC BY-NC-SA 4.0

Files of interest: - images/: standardized handwriting images named {writer_id}_{number:03d}.jpg. - metadata.csv: contains image_path, writer_id, writer_number, original_filename, width, height, file_size_kb.

Preprocessing and standardization

• Converted all images to RGB and saved as JPEG.
• Aspect ratio preserved; images resized using thumbnailing so neither side exceeds 1024 pixels.
• Transparency flattened onto a white background where present.
• Saved with quality=95 to balance fidelity and file size.
• Corrupted or unreadable files were skipped and logged in standardization.log.

Collection, privacy and ethics

Samples were collected from university students in Bangladesh and anonymized by numeric writer IDs. No personal names or identifiable metadata are included. Users must respect the dataset license (CC BY-NC-SA 4.0) and ensure ethical, non-commercial use. If you plan to republish derivative datasets or use the data for commercial purposes, obtain explicit permission from the dataset owner.

Suggested uses

• Writer identification (multi-class classification)
• Writer verification (pairwise or siamese networks)
• Transfer learning and fine-tuning experiments
• Feature extraction and classical machine-learning baselines
• Digital image processing coursework and lab exercises

Recommended practices for experiments

• Use writer_number from metadata.csv as the modeling label.
• For reproducible evaluations, generate or publish explicit stratified train/validation/test splits by writer (not by image) to avoid sample leakage.
• Report per-writer metrics (accuracy, precision, recall) as well as macro-averaged scores to account for imbalance.

Citation

If you use this dataset, please cite:

Handwritten Writer Identification Dataset, standardized from raw writer folders into RGB JPG format. Version 1.0, 91 writers, 424 images.

BibTeX (placeholder):

@dataset{handwriting_writer_identification_dataset,
 title    = {Handwritten Writer Identification Dataset},
 year     = {2026},
 note     = {91 writers, 424 standardized handwriting images},
 howpublished = {Kaggle dataset}
}

Contact and issues

For questions or to report issues, contact the dataset maintainer or open an issue on the dataset page. Include reproducible steps and any relevant original_filename values from metadata.csv when reporting missing or corrupted samples.

This file complements README.md and metadata.csv.

This dataset includes data files and supplementary files related to the review study "The concept of competence in educational research: a knowledge map of main research traditions and topics". It consists of the following documents. Supplementary file 1: Search process. This text document outlines a) our general search strategy, presented in table format, b) the search query we used to retrieve 62 systematic competence reviews from Scopus, c) the search terms employed by those systematic reviews, presented in table format, and d) the search terms we included in our own query, presented in table format. Supplementary files 2a-aj: Retrieved documents. These data files list the 135,298 documents retrieved through the search, along with their corresponding bibliometric information. Supplementary Files 3a-b: VOSviewer map and network files (csv), generated using VOSviewer software, contain results from the direct citation network analysis. Using these files with VOSviewer, our direct citation network graph can be reconstructed in an interactive form. Supplementary File 3c: DCNA cleaned map file (xlsx) provides a list of documents by cluster (research tradition), along with their bibliometric details and related statistics calculated by VOSviewer. With this file, the data can be easily read, filtered and sorted. When we refer to the extensive library of further reading in the article, we are referring to this file. Supplementary file 4: Yearly number of documents in the main traditions of competence research (jpg/png). This supplementary figure illustrates the historical development in the number of documents within each research tradition. It was used to support the writing of the Results section. Supplementary file 5: Data preprocessing algorithm (term co-occurrence analysis). Conducting the term co-occurrence analysis separately for each research tradition presented a key challenge: VOSviewer only accepts files that strictly follow the format of standard database export files. Therefore, we needed to create a separate file for each research tradition, containing all documents assigned to that tradition and formatted to match the structure of a Scopus export file. We addressed this by developing a custom Python code that first extracted cluster assignments from the direct citation network analysis (Supplementary Files 3a-c). It then processed the full dataset (Supplementary File 2), assigning each row to cluster 1, 2, 3, 4, or none, based on matches in DOI, author name, and document title. Supplementary files 6a-l: VOSviewer results files (term co-occurrence analysis). These files are similar to Supplementary Files 3a–c, but pertain to the term co-occurrence analysis instead. For each research tradition, there are three corresponding files: map files, network files, and cleaned map files. The map files (csv) and network files (csv), generated using VOSviewer software, contain results from the term co-occurrence analysis. Using these files with VOSviewer, our term co-occurrence graphs can be recreated in an interactive form. The cleaned map files (xlsx) contain the 500 most frequently occurring terms for each research tradition, along with related statistics calculated by VOSviewer. With these files, the data can be easily read, filtered and sorted.