    TEST DATA SET FOR THE PAPER ``NEER: An Unsupervised Method for
		 Named Entity Evolution Recognition''

These files form the test data set for the paper ``NEER: An
Unsupervised Method for Named Entity Evolution Recognition'' by
N. Tahmasebi et al. They can be used to evaluate systems that detect
named entity changes over time.

The following files are contained in this archive:

    companynames.txt
    locations.txt
    persons.txt

	temporal co-references for named entities from the classes
	company names, locations and persons. There are a total of 86
	co-references, 42 direct and 44 indirect. These correspond to
	33 distinct entities.

	The format is described below in section 1 ``Co-Reference File
	Format''.

    change_years.csv

	Known periods of change for the given entities. The format of
	this file is described below in section 2 ``Changes File
	Format''.

1. Co-Reference File Format
======================

Each file consists of blocks of co-references for the same query
term. Each block is separated by a blank line.

Co-reference lines have the following format:

query term \t co-reference \t class \n

i.e. each column is separated by a TAB character. The class column has
one of the following values:

     2 	  for temporal co-references, i.e. names that are used only for
     	  some periods of the dataset

     1 	  for additional terms used to refer to the entitiy,
          e.g. `musician'


2. Changes File Format
======================

The file changes.csv gives the known change periods (given as the year
of change) for the tested entities. Each line has the format:

query term ; year \n

Note that there may be several change periods for the same entity.


3. Using This Data Set
======================

You are allowed to use this data set to develop, test, and evaluate
your own systems. If you publish your results, please cite this data
set as

    N. Tahmasebi et al.: NEER: An Unsupervised Method for Named Entity Evolution Recognition. In: COLING 2012, pp XXX--XXX.

and send us an email at <tahmasebi@l3s.de>.

If you have any questions or want to suggest any additional queries,
please contact us at <tahmasebi@l3s.de>.
