Parallel text extraction from Global Voices

Global Voices is a community of bloggers and citizen journalists writing and translating news articles in several languages. We crawl articles from this website to create parallel corpora for low-resource languages.

Requirements

Python 2.7 is required. Install dependencies with:

pip install -r requirements.txt

Step 1 (up to 1 week): crawl articles from the website

We can obtain seed URLs from the RSS feeds. It is recommended to add more seeds at different starting dates if starting to crawl from scratch. The crawler will follow the previous and next article links on each page.

For example, to crawl the Malagasy version of the website, run:

mkdir crawl-mg
curl https://mg.globalvoices.org/feed/ | python gv-crawl/make_seeds.py > crawl-mg/seeds.txt
python gv-crawl/crawler.py crawl-mg/seeds.txt crawl-mg --delay 1 2> crawl-mg/crawl.log
python gv-crawl/warc2db.py ./crawl-mg/scrapy.*.warc.gz articles.db

The crawling can be interrupted and restarted; it should resume operation automatically. This also makes incremental crawling possible.

Compressed WARC files containing the crawled pages are created.

Step 1b: crawl English translations

To crawl the English translations

python gv-crawl/en-translation-urls.py en mg articles.db > crawl-mg/en-urls.txt
python gv-crawl/nolink-crawler.py crawl-mg/en-urls.txt crawl-mg --delay 1 2> crawl-mg/en-crawl.log

Step 2: create article database again

After we have crawled several versions of the website, we can find parallel documents for a pair of languages. Fist, we insert all articles in a database (repeat for all languages):

python gv-crawl/warc2db.py ./crawl-mg/scrapy.*.warc.gz articles.db

Step 3: sentence alignment

Then, we use the Gargantua sentence aligner to align the sentences from parallel articles:

export GARGANTUA=/path/to/gargantua
python gv-crawl/db2bidoc.py en mg articles.db $GARGANTUA
cd $GARGANTUA && mkdir corpus_data && cd corpus_data && ../src/sentence-aligner

The db2bidoc.py creates the tokenized/untokenized/info files necessary for Gargantua to run in the $GARGANTUA directory. These intermediary files (corpus_data, corpus_to_align, input_documents) can be deleted after the last step has been run.

Step 4: create aligned XML bitext

Finally, an XML file containing the bitext is created:

python gv-crawl/align2xml.py eng mlg\
	$GARGANTUA/corpus_to_align/align_info.txt\
	$GARGANTUA/corpus_data/output_data_aligned > en-mg.xml

This XML file can be processed with the teny tools to produce parallel text files.

License

Copyright (c) 2013, Victor Chahuneau

All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

• Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
• Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.