Creates knowledge graph from information processed by "Entity Extraction and Linking" module, and "Emotion Recognition from Text" module
Switch branches/tags
Nothing to show
Clone or download
Andrejs Abele
Andrejs Abele added Docker script
6
Latest commit 418e58d May 16, 2017
Permalink
Failed to load latest commit information.
DBpedia_uncompressed added rule to ignore nt files Apr 9, 2017
.gitignore added rule to ignore nt files Apr 9, 2017
Dockerfile
LICENSE Initial commit Apr 9, 2017
README.md added reqired fields Apr 10, 2017
knowledgeGraphCreator.py First version of knowledge Graph Apr 9, 2017
originalConfig.json First version of knowledge Graph Apr 9, 2017

README.md

MixedEmotions' knowledge-graph

Description

Creates Knowledge Graph from information processed by "Entity Extraction and Linking", and "Emotion Recognition from Text" module

This MixedEmotions Knowledge Graph was developed by NUIG.

Knowledge Graph provides insight into relations between recognized entities using semantic knowledge from DBpedia. KG module uses entities that are recognised by "Entity Extraction and Linking" module, and extracts relations between the entities from DBpedia. Once the relations are extracted and filtered, they are stored in Elasticsearch database, where using Kibi they are visualized.

REQUIREMENTS

This package requires python3.5.

Python libraries:

  • elasticsearch
  • Flask

DBpedia dumps from http://wiki.dbpedia.org/downloads-2016-04

Minimal list of required files:

  • infobox_properties_en.ttl
  • instance_types_en.ttl
  • persondata_en.ttl

Elasticsearch: 2.4.1

Kibi: kibi-enterprise-standard-4.6.3-2

USAGE

Type http://localhost:5000/

Description API call
Check default configuration GET /configuration
Modify the configuration POST /configuration
Reset back to default configuration GET /reset
Get status of the module GET /status
Create the Knowledge Graph GET /start

Usage example

Kibi already has to contain the source index pattern. In our example trump_tweets.

trump_tweets: index on elasticsearch that contains data processed by Entity Extraction and Linking, [Emotion Recognition from Text] module, and has a field text (that contains original text which was processed).

From Entity Extraction and Linking we use fields:

  • entity_linking.URI
  • entity_linking.EntityType

From Emotion Recognition from Text we use fields:

  • emotions.emotion

Submit credentials

http://localhost:5000/configuration PUT

{
"credentials": {
"elasticHost": "localhost",
"elasticPassword": "changeme",
"elasticPort": 9220,
"elasticUsername": "elastic"
},
  "variables": {
"inputIndexName": "trump_tweets",
"inputIndexType": "text_review"
}

}

Start Creation of graph

http://localhost:5000/start GET

Monitor progress

http://localhost:5000/status GET

View graph in Kibi

http://localhost:5606/app/kibana#/dashboard/Graph

ACKNOWLEDGEMENT

This development has been partially funded by the European Union through the MixedEmotions Project (project number H2020 655632), as part of the RIA ICT 15 Big data and Open Data Innovation and take-up programme.

MixedEmotions

EU

http://ec.europa.eu/research/participants/portal/desktop/en/opportunities/index.html