Forked for universal push rights!
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 1 commit ahead, 3 commits behind fabricehong:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


We have created an easily re-usable data set of transcriptions of all parliamentary interventions since 1990. Download the CSV or JSON, or play with the data interactively by heading over to:

Play around with search terms, visualizations, and dashboards. Try searching for Raclette for example:


Then make a visualization: go to Visualizations -> Pie Chart -> From New Search -> Split Slices; then Aggregate by Terms, and select your preferred Field.

Raclette by Person

Once you have saved visualizations, you can add them to dashboards. Have a look at this one:

Who talks about cheese?


This is a Elections Hackdays project.
This hackathon is happening Sep. 4-5 2015 in Lausanne (Le Temps newsroom) and Zurich (NZZ newsroom).
A wiki is available for this project.

Where to go next

Do you want to take over this project? Here's what needs to be done first:

  • Clean up the date fields. For now, the script converts to dates to ISO format,
  • it could be better done upstream.
  • Configure elasticsearch to extract significant terms from the transcriptions.


Minutes of the parliament

Parliament API

API doc

How to

Fetching data from parlament website

Fetching the list of URLs for the minutes full text

Source: minutes of National Council from

The script parl_scraper/ fetches the table-of-content HTML files for all minutes, parses the files to extract the fulltext URLs and writes the result to parl_scraper/debates-urls.txt. By default the script's downloading part is commented-out, you can use the pre-downloaded files in parl_scraper/tables-of-content.tar.xz instead.

Scraping the fulltext HTML pages

Install Scrapy:

Project name: "hackaton"

Move to /hackaton

In the parl_scraper directory, run:

scrapy crawl CNbasic2 -o ../data/items-full-final.json
scrapy crawl CNbasic2 -o ../data/items-full-final.csv

Fetching the biographic data for each member of parliament

The script biography_retrieval/retrieve.R downloads all the biographic data into the bio_json and bio_xml subdirectories.

Merging the fulltext and biographic data

The scripts/ script can be used to generate either one JSON file with the merged data, or one JSON file per legislative session.

For the first case, give it the input and output files:

scripts/ data/items-full-final.json data/items-full-final-with-bio.json

For the second case, give it an existing directory as second argument:

scripts/ data/items-full-final.json data/with-bio-split-json

Converting the JSON files to CSV

The program scripts/csv_convert.go can be used to convert a JSON file to CSV. To build, run go build csv_convert.go in the scripts directory.

To run, give it the input and output file names:

scripts/csv_convert data/items-full-final-with-bio.json data/items-full-final-with-bio.csv

To convert all JSON files from data/with-bio-split-json into CSV files in the (pre-existing) directory data/with-bio-split-csv, you can run the following shell command:

for i in data/with-bio-split-json/*.json; do scripts/csv_convert "$i" "${i//json/csv}"; done

Data format:

The data folder contains two representations of the intervention transcripts. The first is a JSON file. It contains an array of objects, of which each object contains the following keys:

Link_subject: link to the page of the subject under discussion
Surname: of the person speaking
Description: of the subject under discussion
Bio: link to the page of the person speaking
Canton: of provenance
Subject_id: of the subject under discussion
Date: of the intervention (DD.MM.YY)
Group: political group of the person speaking at the moment of the intervention
Session_title: title of the session (Séance)
Data: transcript of the intervention
Name: of the person speaking

The second representation is a CSV file with the same data, where each object described above occupies one line.

These data are available in the archives data/ and data/ respectively.

The same datasets split by legislative session are available in the archives data/ and

Python development

Setup virtualenv:

virtualenv env
source env/bin/activate

Install the packages and its dependencies:

python install
python develop

Overview of the architecture

  • Meeting minutes are scraped from confederation website using python scripts in the project parl_scraper
  • We use elasticsearch to expose a text query interface on top of the json's
  • We use Kibana on top of elastic search to do visualization of the text searches

Running docker elasticsearch/kibana containers:

start :


stop :


connect in ssh to container :

./conn [es | kib]


The commands that actually worked for us on DigitalOcean:

docker run -p 3333:9200 -p 3334:9300 -d --name=elasticsearch elasticsearch
docker run --link elasticsearch:elasticsearch -p 3335:5601 -d kibana

Import into elasticsearch

To import the data into elasticsearch, unzip the JSON data and move all JSON files to a new folder:

cd data
cd ..

Then use the upload script to upload to the elasticsearch server. Assuming the server is running on localhost on port 3333, the following command uploads all the entries in all the json files to elasticsearch:

python scripts/ data/with-bio-split-json http://localhost:3333/parlament/intervention

This creates the entries under the parlament index.


Once the data is uploaded into elasticsearch, go to Kibana (at localhost:3335 if you followed the instructions above), then create a default index:

Index config

Now the data is searchable. Refer to the tutorial. To change the default date range, you need to change Kibana's source code. Connect to your kibana instance, install vim, and change a line in the public index.js file:

host~$ docker exec -i -t kibana bash
kibana~$ apt-get update && apt-get install vim
kibana~$ vim /opt/kibana/src/public/index.js

Find the following block:

 var timeDefaults = {
    from: 'now-15m',
    to: 'now',
    mode: 'quick'

And replace it with:

 var timeDefaults = {
    from: 'now-25y',
    to: 'now',
    mode: 'quick'

Now reload Kibana in your browser.