DataShare aims at allowing for valuable knowledge about people and companies locked within hundreds of pages of documents inside a computer to be sieved into indexes and shared securely within a network of trusted individuals, fostering unforeseen collaboration and prompting new and better investigations that uncover corruption, transnational crime and abuse of power.

DataShare: connecting local data with a global collective intelligence

An Extensible Multilingual Information Extraction and Search Platform

• Extract Text from Files;
• Extract Organizations, Persons and Locations from Text;
• Index and Search all

Multithreaded and Distributed Processings

Local or Remote Indexing

You can download the script datashare.sh and execute it. It will :

• download redis, elasticsearch and datashare docker containers
• initialize an elasticsearch index with datashare mapping
• provide CLI to run datashare extract, index, name finding tasks
• provide a WEB GUI to run datashare extract, index, name finding tasks, and search in the documents

To access web GUI, go in your documents folder and launch path/to/datashare.sh -w then connect datashare on http://localhost:8080

If you want to avoid synchronization of NLP models (offline use) then do export DS_JAVA_OPTS="-DDS_SYNC_NLP_MODELS=false" before launching the datashare.sh script.

You can use the datashare docker container only for HTTP exposed name finding API.

Just run :

docker run -ti -p 8080:8080 -v /path/to/dist/:/home/datashare/dist icij/datashare:0.10 -m NER -w

A bit of explanation :

• -w tells datashare to run the webserver. It is launched on 8080 that's why the port is mapped for docker
• -m NER runs datashare without index at all on a stateless mode
• -v /path/to/dist:/home/datashare/dist maps the directory where the NLP models will be read (and downloaded if they don't exist)

Then query with curl the server with :

curl -i localhost:8080/ner/findNames/CORENLP --data-binary @path/to/a/file.txt

The last path part (CORENLP) is the framework. You can choose it among CORENLP, IXAPIPE, MITIE or OPENNLP.

Implementations

• TikaDocument from ICIJ/extract

  Apache Tika v1.18 (Apache Licence v2.0)

  with Tesseract v4.0 alpha

Support

Tika File Formats

Implementations

• org.icij.datashare.text.nlp.corenlp.CorenlpPipeline

  Stanford CoreNLP v3.8.0, (Conditional Random Fields), Composite GPL v3+

• org.icij.datashare.text.nlp.ixapipe.IxapipePipeline

  Ixa Pipes Nerc v1.6.1, (Perceptron), Apache Licence v2.0

• org.icij.datashare.text.nlp.mitie.MitiePipeline

  MIT Information Extraction v0.8, (Structural Support Vector Machines), Boost Software License v1.0

• org.icij.datashare.text.nlp.opennlp.OpennlpPipeline

  Apache OpenNLP v1.6.0, (Maximum Entropy), Apache Licence v2.0

Natural Language Processing Stages Support

Named Entity Recognition Language Support

Named Entity Categories Support

Parts-of-Speech Language Support

Implementations

• org.icij.datashare.text.indexing.elasticsearch.ElasticsearchIndexer

  Elasticsearch v6.1.0, Apache Licence v2.0

Requires JDK 8, Maven 3

From datashare root directory, type: mvn package

DataShare is released under the GNU Affero General Public License

We welcome feedback as well as contributions!

For any bug, question, comment or (pull) request,

please contact us at [email protected]

• Data Sharing module

  • Networking module

  • Content Management module

  • User Management module

  • Request and Exchange Protocol