Size Categories:
Language Creators:
Annotations Creators:
Source Datasets:
usually , he would be tearing around the living room , playing with his toys .
but just one look at a minion sent him practically catatonic .
that had been megan 's plan when she got him dressed earlier .
he 'd seen the movie almost by mistake , considering he was a little young for the pg cartoon , but with older cousins , along with her brothers , mason was often exposed to things that were older .
she liked to think being surrounded by adults and older kids was one reason why he was a such a good talker for his age .
`` are n't you being a good boy ? ''
she said .
mason barely acknowledged her .
instead , his baby blues remained focused on the television .
since the movie was almost over , megan knew she better slip into the bedroom and finish getting ready .
each time she looked into mason 's face , she was grateful that he looked nothing like his father .
his platinum blond hair and blue eyes were completely hers .
it was only his build that he was taking after his father .
where megan was a diminutive 5'3 '' , davis was 6'1 '' and two hundred pounds .
mason was already registering off the charts in height and weight according to his pediatrician .
davis had seen mason only twice in his lifetime-the day he had been born and the day he came home from the hospital .
after that , he had n't been interested in any of the pictures and emails megan sent .
with his professional football career on the rise , davis had n't wanted to be shackled with the responsibilities of a baby .
instead , he wanted to spend his time off the field partying until all hours of the night .
he only paid child support when megan threatened to have his wages garnished .
she dreaded the day when mason was old enough to ask about his father .
she never wanted anything in the world to hurt him , and she knew that being rejected by his father would .
with a sigh , she stepped into the dress and slid it over her hips .
wrestling around to get the zipper all the way up caused her to huff and puff .
standing back from the mirror , she turned to and fro to take in her appearance .
she 'd always loved how the dress made her feel sexy , but at the same time was very respectable .
while it boasted a sweetheart neckline , the hemline fell just below her knees .
she put on her pearls-a high school graduation gift from her uncle aidan , or `` ankle '' , as she often called him .
aidan was her mother 's baby brother and only son of the family .
when she was born , he was only eight and a half .
as the first grandchild , megan spent a lot of time with her grandparents , and that in turn , meant she spent a lot of time with aidan .
he had devoted hours to holding her and spoiling her rotten .
when it came time for her to talk , she just could n't seem to get `` uncle aidan '' out .
instead , she called him `` ankle . ''
it was a nickname that had stuck with him even now that he was thirty-four and married .
while it had been no question that she wanted him as godfather for mason , she had been extremely honored when he and his wife , emma , had asked her to be their son , noah 's , godmother .
she loved her newest cousin very much and planned to be the best godmother she could for him .
as she stepped out of the bedroom , she found that mason had yet to move .
`` okay buddy , time to go . ''
when he started to whine , she shook her head .
`` we have such a fun day ahead of us .
it 's noah 's baptism , and then there 's a party at uncle aidan and aunt emma 's house . ''
`` beau ? ''
he asked .
she laughed .
`` yes , you 'll get to see and play with beau , too . ''
as she went to the couch and picked him up , she could n't help finding it amusing that out of everyone he was going to see today , he was most excited about being with aidan and emma 's black lab , beau .
one day when they had their own place again , she would get him a dog .
he loved them too much to be denied .
`` oomph , '' she muttered , as they started up the basement stairs .
`` heawy ? ''
he asked .
`` yes , you 're getting to be such a big , heavy boy . ''
when they made it to the kitchen , megan paused to catch her breath .
she only had a second before her mother breezed in with sean , and her youngest brother , gavin .
`` ready ? ''
she asked .
megan nodded .
feeling like she was once again a teenager , she filed behind her parents as they headed into the garage .
`` i want to drive , '' gavin said .
with a smirk , sean replied , `` like i 'm gon na let you drive my car . ''
he then slid into the driver 's seat as gavin reluctantly walked around to the passenger 's side .
`` we 'll see you there in just a few , '' her mother called .
sean acknowledged her with a two finger salute before cranking up and pulling down the driveway .
megan worked to get mason into the car seat in her parents ' land rover .
once he was safely strapped and buckled in , she hopped in beside him .
her parents rattled along to each other as they made their way through the tree-lined suburbs where megan had grown up .
while some might look on her as having a mark against her character being an unwed mother , she had lived a relatively non-rebellious life .
even though she 'd been a cheerleader and ran with the popular crowd in school , she rarely partied to excess .
instead , she had focused on getting good grades .
at that time , she had her heart set on going to medical school and becoming a doctor .
from the time she was a little girl , she had wanted nothing more than to help people .
she was always mending birds with broken wings or trying to resuscitate squirrels who had been hit by cars .
she ditched playing princess for playing `` hospital . ''
her desire to become a doctor was why she needed the best scores and best activities and why she generally shunned any temptations to lead her off the right path .
she had even managed to bypass the usual freshman craziness when she went off to the university of georgia .
it was n't until she fell in love for the first time in her life that she threw everything away .
sadly , she could n't say that her first love was davis , mason 's father .
instead , it was another football player , this time a running back at uga , who captured and later broke her heart a year later .
carsyn ran with the fast crowd , and when she was with him , she partied and drank too much .
he was controlling and possessive , and he wanted all of her time .
when she was with him , she had little time for studying .
with her grades already in the toilet , she was unprepared for the emotional breakdown she experienced when carsyn broke up with her .
devastated , she stopped going to class and ended up flunking the semester .
by the time she got back on track with her grades , she had abandoned any hope of medical school .
instead , she decided that she would become a nurse , which would fulfill her need to care for sick people .
of course , her relationship with davis ended up derailing shortly before graduation when she got pregnant unexpectedly .
she had to take several semesters off after mason was born .
she was a few years off from when she had originally planned on graduating , but she was excited after everything had that had happened , she was finally finishing .
her mother 's voice brought megan out of her thoughts .
`` here we are , '' she said pleasantly .
leaning forward in her seat , megan eyed the clock on the dashboard .
she was n't surprised to see they had arrived half an hour before the baptism started .
one thing her mother prided herself on was being on time and lending a hand .
as they started into the church , her mother reached for mason .
`` we 'll take him so you can go see if emma needs any help . ''
megan bent over to kiss mason 's cheek .
`` see you in a little while , sweetie . ''
he grinned and then happily dodged her mother 's arms for her father 's instead , which made megan smile .
he was such a man 's man already .

Dataset Card for BookCorpus

Dataset Summary

Books are a rich source of both fine-grained information, how a character, an object or a scene looks like, as well as high-level semantics, what someone is thinking, feeling and how these states evolve through a story.This work aims to align books to their movie releases in order to providerich descriptive explanations for visual content that go semantically farbeyond the captions available in current datasets.

Supported Tasks and Leaderboards

More Information Needed


More Information Needed

Dataset Structure

Data Instances

In the original dataset described by Zhu and Kiros et al., BookCorpus contained 11,038 books. However, based on the files obtained, there appear to be only 7,185 unique books (excluding romance-all.txtand adventure-all.txt as explained in 2.2.1). Potential duplicates were identified based on file names, which suggested that 2,930 books may be duplicated. Using the diff Unix program, it was confirmed that BookCorpus contained duplicate, identical text files for all but five of these books. The five exceptions were manually inspected:

  • 299560.txt (Third Eye Patch), for which slightly different versions appeared in the 鈥淭hriller鈥 and 鈥淪cience Fiction鈥 genre folders (only 30 lines differed)
  • 529220.txt (On the Rocks), for which slightly different versions appeared in the 鈥淟iterature鈥 and 鈥淪cience Fiction鈥 genre folders (only the title format differed)
  • Hopeless-1.txt, for which identical versions appeared in the 鈥淣ew Adult鈥 and 鈥淵oung Adult鈥 genre folders, and a truncated version appeared in the 鈥淩omance鈥 folder (containing 30% of the full word count)
  • u4622.txt, for which identical versions appeared in the 鈥淩omance鈥 and 鈥淵oung Adult鈥 genre folders, and a slightly different version appeared in the 鈥淪cience Fiction鈥 folder (only 15 added lines)
  • u4899.txt, for which a full version appeared in the 鈥淵oung Adult鈥 folder and a truncated version (containing the first 28 words) appeared in the 鈥淪cience Fiction鈥 folder

Combined with the diff results, the manual inspection confirmed that each filename represents one unique book, thus BookCorpus contained at most 7,185 unique books.


  • Size of downloaded dataset files: 1.18 GB
  • Size of the generated dataset: 4.85 GB
  • Total amount of disk used: 6.03 GB

An example of 'train' looks as follows.

    "text": "But I traded all my life for some lovin' and some gold"

Data Fields

Each book in BookCorpus simply includes the full text from the ebook (often including preamble, copyright text, etc.). However, in research that BookCorpus, authors have applied a range of different encoding schemes that change the definition of an 鈥渋nstance鈥 (e.g. in GPT-N training, text is encoded using byte-pair encoding). The data fields are the same among all splits. There is no label or target associated with each instance (book). The text from each book was originally used for unsupervised training by Zhu and Kiros et al., and the only label-like attribute is the genre associated with each book, which is provided by Smashwords. No relationships between individual instances (books) are made explicit. Grouped into folders by genre, the data implicitly links books in the same genre. It was found that duplicate books are implicitly linked through identical filenames. However, no other relationships are made explicit, such as books by the same author, books in the same series, books set in the same context, books addressing the same event, and/or books using the same characters.


  • text: a string feature.

Data Splits

There are no recommended data splits. The authors use all books in the dataset for unsupervised training, with no splits or subsamples.

name train
plain_text 74004228

Dataset Creation

Curation Rationale

The books in BookCorpus were self-published by authors on, likely with a range of motivations. While we can safely assume that authors publishing free books via had some motivation to share creative works with the world, there is no way to verify they were interested in training AI systems. For example, many authors in BookCorpus explicitly license their books 鈥渇or [the reader鈥檚] personal enjoyment only,鈥 limiting reproduction and redistribution. When notified about BookCorpus and its uses, one author from Smashwords said 鈥渋t didn鈥檛 even occur to me that a machine could read my book鈥 [].

Source Data

Initial Data Collection and Normalization

Per Bandy and Vincent (2021), the text for each instance (book) was acquired via download from The data was collected via scraping software. While the original scraping program is not available, replicas (e.g. operate by first scraping to generate a list of links to free ebooks, downloading each ebook as an epub file, then converting each epub file into a plain text file. Books were included in the original Book-Corpus if they were available for free on and longer than 20,000 words, thus representing a non-probabilistic convenience sample. The 20,000 word cutoff likely comes from the Smashwords interface, which provides a filtering tool to only display books 鈥淥ver 20K words.鈥 The individuals involved in collecting BookCorpus and their compensation are unknown. The original paper by Zhu and Kiros et al. ( does not specify which authors collected and processed the data, nor how they were compensated. The timeframe over which BookCorpus was collected is unknown as well. BookCorpus was originally collected some time before the original paper ( was presented at the International Conference on Computer Vision (ICCV) in December 2015. It is unlikely that any ethical review processes were conducted. Zhu and Kiros et al. ( do not mention an Institutional Review Board (IRB) or other ethical review process involved in their original paper. The dataset is related to people because each book is associated with an author (please see the "Personal and Sensitive Information" section for more information on this topic).

Bandy and Vincent also assert that while the original paper by Zhu and Kiros et al. ( did not use labels for supervised learning, each book is labeled with genres. It appears genres are supplied by authors themselves. It is likely that some cleaning was done on the BookCorpus dataset. The .txt files in BookCorpus seem to have been partially cleaned of some preamble text and postscript text, however, Zhu and Kiros et al. ( do not mention the specific cleaning steps. Also, many files still contain some preamble and postscript text, including many sentences about licensing and copyrights. For example, the sentence 鈥減lease do not participate in or encourage piracy of copyrighted materials in violation of the author鈥檚 rights鈥 occurs at least 40 times in the BookCorpus books_in_sentences files. Additionally, based on samples we reviewed from the original BookCorpus, the text appears to have been tokenized to some degree (e.g. contractions are split into two words), though the exact procedure used is unclear. It is unknown if some of the "raw" data was saved in addition to the clean data. While the original software used to clean the BookCorpus dataset is not available, replication attempts provide some software for turning .epub files into .txt files and subsequently cleaning them.

Who are the source language producers?

Per Bandy and Vincent (2021), the data in BookCorpus was produced by self-published authors on and aggregated using scraping software by Zhu and Kiros et al.


Annotation process

More Information Needed

Who are the annotators?

More Information Needed

Personal and Sensitive Information

Per Bandy and Vincent (2021), it is unlikely that authors were notified about data collection from their works. Discussing BookCorpus in 2016, Richard Lea wrote in The Guardian that 鈥淭he only problem is that [researchers] didn鈥檛 ask鈥 ( When notified about BookCorpus and its uses, one author from Smashwords said 鈥渋t didn鈥檛 even occur to me that a machine could read my book鈥 ( Authors did not consent to the collection and use of their books. While authors on published their books for free, they did not consent to including their work in BookCorpus, and many books contain copyright restrictions intended to prevent redistribution. As described by Richard Lea in The Guardian (, many books in BookCorpus include: "a copyright declaration that reserves 鈥渁ll rights鈥, specifies that the ebook is 鈥渓icensed for your personal enjoyment only鈥, and offers the reader thanks for 鈥渞especting the hard work of this author.鈥' Considering these copyright declarations, authors did not explicitly consent to include their work in BookCorpus or related datasets. Using the framework of consentful tech (, a consent- ful version of BookCorpus would ideally involve author consent that is Freely given, Reversible, Informed, Enthusiastic, and Specific (FRIES). It is unlikely that authors were provided with a mechanism to revoke their consent in the future or for certain uses. For example, if an author released a book for free before BookCorpus was collected, then changed the price and/or copyright after BookCorpus was collected, the book likely remained in BookCorpus. In fact, preliminary analysis suggests that this is the case for at least 438 books in BookCorpus which are no longer free to download from Smashwords, and would cost $1,182.21 to purchase as of April 2021.

Considerations for Using the Data

The composition of BookCorpus or the way it was collected and preprocessed/cleaned/labeled might impact future uses. At the very least, the duplicate books and sampling skews should guide any future uses to curate a subsample of BookCorpus to better serve the task at hand. An analysis of the potential impact of BookCorpus and its use on data subjects has not been conducted. Richard Lea interviewed a handful of authors represented in BookCorpus (Richard Lea).

Social Impact of Dataset

The dataset contains data that might be considered sensitive. The aforementioned contact information (email addresses) is sensitive personal information.

Discussion of Biases

BookCorpus contains free books from which are at least 20,000 words long. Based on metrics from Smashwords, 11,038 books (as reported in the original BookCorpus dataset) would have represented approximately 3% of the 336,400 books published on Smashwords as of 2014, while the 7,185 unique books we report would have represented 2%. For reference, as of 2013, the Library of Congress contained 23,592,066 cataloged books (Audrey Fischer).

There are some errors, sources of noise, or redundancies in BookCorpus. While some book files appear to be cleaned of preamble and postscript text, many files still contain this text and various other sources of noise. Of particular concern is that we found many copyright-related sentences, for example:

  • 鈥渋f you鈥檙e reading this book and did not purchase it, or it was not purchased for your use only, then please return to and purchase your own copy.鈥 (n=788)
  • 鈥渢his book remains the copyrighted property of the author, and may not be redistributed to others for commercial or non-commercial purposes...鈥 (n=111)
  • 鈥渁lthough this is a free book, it remains the copyrighted property of the author, and may not be reproduced, copied and distributed for commercial or non-commercial purposes.鈥 (n=109)
  • 鈥渢hank you for respecting the author鈥檚 work鈥 (n=70)
  • 鈥渘o part of this publication may be copied, reproduced in any format, by any means, electronic or otherwise, without prior consent from the copyright owner and publisher of this book鈥 (n=16)

Note that these sentences represent noise and redundancy. As previously noted, BookCorpus also contains many duplicate books: of the 7,185 unique books in the dataset, 2,930 occurred more than once. Most of these (N=2,101) books appeared twice, though many were duplicated multiple times, including some books (N=6) with five copies in BookCorpus. See Table 2.

Other Known Limitations

There are no export controls or other regulatory restrictions that apply to the dataset or to individual instances. Some information is missing from individual instances (books). 98 empty book files were found in the folder downloaded from Zhu and Kiros et al. Also, while the authors collected books longer than 20,000 words, 655 files were shorter than 20,000 words, and 291 were shorter than 10,000 words, suggesting that many book files were significantly truncated from their original text.

There were no ethical review processes conducted. Zhu and Kiros et al. do not mention an Institutional Review Board (IRB) or other ethical review process involved in their original paper. Bandy and Vincent strongly suggest that researchers should use BookCorpus with caution for any task, namely due to potential copyright violations, duplicate books, and sampling skews.

Additional Information

Dataset Curators

More Information Needed

Licensing Information

The books have been crawled from, see their terms of service for more information.

A data sheet for this dataset has also been created and published in Addressing "Documentation Debt" in Machine Learning Research: A Retrospective Datasheet for BookCorpus.

Citation Information

    title = {Aligning Books and Movies: Towards Story-Like Visual Explanations by Watching Movies and Reading Books},
    author = {Zhu, Yukun and Kiros, Ryan and Zemel, Rich and Salakhutdinov, Ruslan and Urtasun, Raquel and Torralba, Antonio and Fidler, Sanja},
    booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
    month = {December},
    year = {2015}


Thanks to @lewtun, @richarddwang, @lhoestq, @thomwolf for adding this dataset.

Downloads last month

Models trained or fine-tuned on bookcorpus