wikipedia-spark

This repo is mainly a wrapper of idio/json-wikipedia, which itself bring some improvement over diegoceccarelli/json-wikipedia.

It combines databricks/spark-xml to read wikipedia dump (without decompressing the whole files :) and the ArticleParser from json-wikipedia to parse wikipedia articles. It brings the speed from Spark and the ability to save easily in parquet or json format, in a distributed maner for instance on S3 :)

build

1. Build json-wikipedia using the command mvn assembly:assembly (for more information see their readme)

Actually you shoud remove the spark dependency from the pom.xml because it's big and is still at version 1.3 ...

Once the json-wikipedia-1.2.0-jar-with-dependencies.jar is build, put it in a lib folder at the root of this repo.

2. Build wikipedia-spark using sbt assembly or sbt docker (which also builds you an image ready to go)

usage

docker run --rm asgard/wikipedia-spark:latest spark-submit /home/wikipedia-spark/wikipedia-spark-assembly-0.0.1.jar ...

Usage: SparkWikipediaParser [options]

  --inputPath <value>
        path to xml dump or folder with xml dumps files
  --outputPath <value>
        path to output file / folder
  --outputFormat <value>
        Format for the ouput: parquet or json
  --lang <value>
        language of the dump
  --test
        Flag to test on a few lines