notnews / archive_news_cc Goto Github PK

The repository provides scripts for downloading the data, and link to two datasets that were built using the scripts:

• Scripts
• Data

Download closed caption transcripts of nearly 1.3M news shows from http://archive.org.

There are three steps to downloading the transcripts:

1. We start by searching https://archive.org/advancedsearch.php with collection collection:"tvarchive". This gets us unique identifiers for each of the news shows. An identifier is a simple string that combines channel_name, show_name, time, and date. The current final list of identifiers (2009--Nov. 2017) is posted here.

2. Next, we use the identifier to build a URL where the metadata file and HTML file with the closed captions is posted. The general base URL is http://archive.org/download followed by the identifier.

3. The third script parses the downloaded metadata and HTML closed caption files and creates a CSV along with the meta data.

For instance, we will go http://archive.org/download/CSPAN_20090604_230000 for identifier CSPAN_20090604_230000 And from http://archive.org/download/CSPAN_20090604_230000/CSPAN_20090604_230000_meta.xml, we read the link http://archive.org/details/CSPAN_20090604_230000, from which we get the text from HTML file. We also store the meta data from the META XML file.

1. Get Show Identifiers

   • Get Identifiers For Each Show (Channel, Show, Date, Time)
   • Produces data/search.csv

2. Download Metadata and HTML Files

   • Download the Metadata and HTML Files
   • Saves the metadata and HTML files to two separate folders specified in --meta and --html respectively. The default folder names are meta and html respectively.

3. Parse Metadata and HTML Files

   • Parses metadata and HTML Files and Saves to a CSV
   • Produces a CSV. Here's an example

1. Get all TV Archive identifiers from archive.org.

   python get_news_identifiers.py -o ../data/search.csv
   

2. Download metadata and HTML files for all the shows in the sample input file

   python scrape_archive_org.py ../data/search-test.csv
   

   This will create two directories meta and html by default in the same folder as where the script is. We have included the first 25 metadata and first 25 html files.

   You can change the folder for meta by using the --meta flag. To change the directory for html, use the --html flag and specify the new directory. For instance,

   python scrape_archive_org.py --meta meta-foxnews --html html-foxnews ../data/search-test.csv
   

   Use -c/--compress option to store and parse the downloaded files in compression format (GZip).

3. Parse and extract meta fields and text from sample metadata and HTML files.

   python parse_archive.py ../data/search-test.csv
   

   A sample output file.

The data are hosted on Harvard Dataverse

Dataset Summary:

1. 500k Dataset from 2014:

   • CSV: archive-cc-2014.csv.xza* (2.7 GB, split into 2GB files)
   • HTML: html-2014.7za* (10.4 GB, split into 2GB files)

2. 860k Dataset from 2017:

   • CSV: archive-cc-2017.csv.gza* (10.6 GB, split into 2GB files)
   • HTML: html-2017.tar.gza* (20.2 GB, split into 2GB files)
   • Meta: meta-2017.tar.gza* (2.6 GB, split into 2GB files)

3. 917k Dataset from 2022:

   • CSV: archive-cc-2022.csv.gza* (12.6 GB, split into 2GB files)
   • HTML: html-2022.tar.gza* (41.1 GB, split into 2GB files)
   • Meta: meta-2022.tar.gz (2.1 GB)

4. 179k Dataset from 2023:

   • CSV: archive-cc-2023.csv.gz (1.7 GB)
   • HTML: html-2023.tar.gza* (7.3 GB, split into 2GB files)
   • Meta: meta-2023.tar.gz (317 MB)

Please note that the file sizes and splitting information mentioned above are approximate.

We are releasing the scripts under the MIT License.

Please credit Internet Archive for the data.

If you wanted to refer to this particular corpus so that the research is reproducible, you can cite it as:

archive.org TV News Closed Caption Corpus. Laohaprapanon, Suriyan and Gaurav Sood. 2017. https://github.com/notnews/archive_news_cc/