download the dataset

[YuMiaoTHU]

thanks for your excellent work！

when I run download-bbc-articles.py, it showed that

I want to konw why, thanks for your help~

[shashiongithub]

Maybe accessing WebArxive urls are restricted! If the problem remains, drop me an email.

[YuMiaoTHU]

Thanks for your reply! sometimes It' hard for us to access some website....... I still can't get the dataset, could you send me the raw data or the processed data via google drive or dropbox? Thanks for your hard work!

[thinkwee]

I have the same problem. The server is not stable. I downloaded about 2000 data for the first time then i rerun the scripts, it cannot download anymore.

[joelowj]

Hi @shashiongithub, I am having similar issues downloading the data with the script. At the moment, I am working on a paper and would love to use xsum dataset for my experiment. I was hoping if you could share them with me through other channels. I tried contacting you through your email but could not get the email to send to your mailbox. My email is [email protected] Thanks a lot!

[shashiongithub]

Here is the dataset:

http://kinloch.inf.ed.ac.uk/public/XSUM-EMNLP18-Summary-Data-Original.tar.gz

Please use train, development and test ids from github to split into subsets.
Let me know if you have any questions.

[shahbazsyed]

Hi,
The link provided above is broken. Is there another way to get the dataset ?

[mingzi151]

Hey, I'm not able to open the link either. Can you please help?

[shashiongithub]

http://kinloch.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz

[shahbazsyed]

Thanks!

[fatihbeyhan]

hey! link is broken :/ can you share updated one for me, so i can download the dataset..

[shashiongithub]

http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz

[sshleifer]

That url creates a dir called bbc-summary-data containing files like bbc-summary-data/{bbcid}.summary.
Which code is meant to be run after that to continue preprocessing? bbcid.summary files are not mentioned in the README. Thanks!

[sshleifer]

First file bbc-summary-data/10000983.summary looks like this:

[shashiongithub]

Few things to keep in mind:

1. There are some extra summary files here, you should ignore them
   (they have more one sentence in their summary etc).

Please use the training/dev/test ids provided here to find which one to use:
https://github.com/EdinburghNLP/XSum/tree/master/XSum-Dataset

2. There are few mismatches (between data here and the formats on the Github):

bbc-summary-data/bbcid.summary --> xsum-extracts-from-downloads/bbcid.data

In each summary file:
[SN]URL[SN] => [XSUM]URL[XSUM]
[SN]TITLE[SN] => Ignore this, not used.
[SN]FIRST-SENTENCE[SN] => [XSUM]FIRST-SENTENCE[XSUM]
[SN]RESTBODY[SN] => [XSUM]RESTBODY[XSUM]

With these changes the preprocessing scripts should work.

[sshleifer]

1. Verifying that I don't need to run prepare_bbc_data.py after doing the SN --> XSUM replacement, right?

2. Which field is the summary? Or is that in another file?

For context, I'm trying to replicate the results in the bart paper

Thanks!

[msadat3]

Hello,

I am also not being able to access any links posted in this thread. Could you please post a working URL?

Update: I found that the posted url works if we do "wget http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz" from a terminal. It does not work from my browser.

[StevenTang1998]

Hello,

I am also not being able to access any links posted in this thread. Could you please post a working URL?

Update: I found that the posted url works if we do "wget http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz" from a terminal. It does not work from my browser.

The same problem here. Could you please post a new URL?

[anamtaamin]

hey! link is broken :/ can you share updated one for me, so i can download the dataset..
try...
https://huggingface.co/datasets/xsum/resolve/main/data/XSUM-EMNLP18-Summary-Data-Original.tar.gz

[mikechen66]

It geneates an error if use the downloaded dataset. Please see the details as follows.

While write the abiove-mentioned weblink (listed as follows again)

http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz

into the xsum.py

_URL_DATA = "http://bollin.inf.ed.ac.uk/public/direct/XSUM-EMNLP18-Summary-Data-Original.tar.gz"

And then run the code as follows.

from datasets import load_dataset
raw_datasets = load_dataset("xsum.py",  "raw_datasets")

It generates the error as follows..

ReadError: unexpected end of data

The above exception was the direct cause of the following exception:
File ~/miniconda3/envs/tf/lib/python3.10/site-packages/datasets/builder.py:1712, in GeneratorBasedBuilder._prepare_split_single(self, gen_kwargs, fpath, file_format, max_shard_size, split_info, check_duplicate_keys, job_id)
1710 if isinstance(e, SchemaInferenceError) and e.context is not None:
1711 e = e.context
-> 1712 raise DatasetGenerationError("An error occurred while generating the dataset") from e
1714 yield job_id, True, (total_num_examples, total_num_bytes, writer._features, num_shards, shard_lengths)

DatasetGenerationError: An error occurred while generating the dataset

The dataset source may have a problem.

Notes:

However, if use the original code, it can run successfully.

from datasets import load_dataset
raw_datasets = load_dataset("xsum", split="train")