Examples for the IPTC EXTRA classification engine. This repository contains the objects/resources that can be used to validate the proper functionality of IPTC EXTRA project.

See the other repositories of the IPTC EXTRA project:

• https://github.com/iptc/extra-core
• https://github.com/iptc/extra-ext

There are five types of resources: rules, schemas, taxonomies, topics and corpora of documents. After a successful deployment of EXTRA Toolkit, following the guide you can find in the extra-ext repository, these objects should be inserted into the platform.

Note! You can import all the resources by running this python script.

$ python insert_resources.py params.json

The only change you need to make is to add the directory of the documents (documents_file) in the params.json file. To get the document please fill the form you can find here.

In the current version of EXTRA, there are two taxonomies with their corresponding topics. More specifically, the two taxonomies are IPTC's Media Topics in english and german respectively.

To create these taxonomies in EXTRA use the corresponding EXTRA API method:

POST /taxonomies

{
  "language": "english",
  "name": "IPTC Media Topics"
}

POST /taxonomies

{
  "language": "german",
  "name": "IPTC Media Topics"
}

Upon successful creation, each of these methods will return the newly created taxonomy with a unique taxonomy id. For example:

Response

201 - Created

{
  "id": "5901b9e5c41479000146ced2",
  "language": "english",
  "name": "IPTC Media Topics"
}

To create topics in a taxonomy e.g. in the taxonomy with id=5901b9e5c41479000146ced2 created before:

POST /taxonomies/5901b9e5c41479000146ced2/topics

{
  "taxonomyId": "5901b9e5c41479000146ced2",
  "name": "arts, culture and entertainment",
  "definition": "Matters pertaining to the advancement and refinement of the human mind, of interests, skills, tastes and emotions ",
  "topicId": "medtop:01000000"
}

To insert these two taxonomies and their corresponding topics use the script files and the guide can be found here.

To create a new schema in EXTRA:

POST /schemas

{
  "name":"Apa Schema",
  "fields":[
    {"name":"title", "hasParagraphs":false, "hasSentences":true, "textual":true},
    {"name":"subtitle", "hasParagraphs":false, "hasSentences":true, "textual":true},
    {"name":"body", "hasParagraphs":true, "hasSentences":true, "textual":true},
    {"name":"id", "hasParagraphs":false, "hasSentences":false, "textual":false},
    {"name":"slugline", "hasParagraphs":false, "hasSentences":false, "textual":false},
    {"name":"versionCreated", "hasParagraphs":false, "hasSentences":false, "textual":false}],
    "language":"german"
}

In EXTRA there are two schemas one for articles from Thomson Reuters and one for Austrian Press Agency. You can see these schemas here.

To create a new corpus that follows a specific schema and its documents are annotated with the topics of a specific taxony:

POST /corpora

{
  "name": "Apa Corpus",
  "language": "german",
  "schemaId": "591f070c30c49e00011de8eb" ,
  "taxonomyId": "5901b9ebc41479000146ced3"
}

To index documents, for developement and testing of rules, into the newly created corpus, follow the guide here.

To create a new rule for a given taxonomy and a topic within that taxonomy:

POST /rules

{
	"name": "Test Rule",
	"query": "(or (and (title adj/regexp \"\\d+\\-?\\s+?year\\-?\\s?old\") (body any/stemming \"boy child children girl infant juvenile kid newborn schoolboy schoolgirl toddler\") ) (and (title adj/regexp \"\\d+\\-?\\s+?month\\-?\\s?old\") (body any/stemming \"boy child children girl infant juvenile kid newborn schoolboy schoolgirl toddler\") ) )",
	"taxonomy": "5901b9e5c41479000146ced2",
	"topicId": "medtop:20000790"
}

The taxonomy id and the topic id must correspond to resources that have already created into the platform. The rules developed during EXTRA project can be found here. To insert them into EXTRA use the script and follow the corresponding guide here.

Technical details: Manos Schinas ([email protected]), Symeon Papadopoulos ([email protected])

Project as a whole: IPTC ([email protected])