Source code for capreolus.collection.nyt

import json

from capreolus import constants
from capreolus.utils.loginit import get_logger

from . import Collection, IRDCollection

[docs]logger = get_logger(__name__)
[docs]PACKAGE_PATH = constants["PACKAGE_PATH"]


@Collection.register
[docs]class NYT(IRDCollection):
    """New York Times collection. See https://catalog.ldc.upenn.edu/LDC2008T19"""

[docs]    module_name = "nyt"
[docs]    ird_dataset_name = "nyt"
[docs]    collection_type = "JsonCollection"

[docs]    def doc_as_json(self, doc):
        content = " ".join((doc.headline, doc.body))
        return json.dumps({"id": doc.doc_id, "contents": content})