.. _fwflag_add_ner:
==========
--add_ner
==========
Switch
======

--add_ner

Description
===========

Extracts named-entity features from a directory of Stanford CoreNLP semantic-parse XML files.

Argument and Default Value
==========================

No argument. Boolean switch, default: ``False``.

Details
=======

Not a database-driven feature at all — ``SemanticsExtractor.addNERTable()`` is meant to read gzipped Stanford CoreNLP XML output directly off disk (one subdirectory of files per group) rather than from any message/feature table, using a corpus directory path.

**Known issue — always crashes, regardless of setup.** The code that constructs the ``SemanticsExtractor`` passes ``args.corpdir`` as that directory path, but no ``--corpdir`` (or any other name) flag registers a ``corpdir`` destination anywhere in ``dlatkInterface.py``. Verified: running ``--add_ner`` on its own fails immediately with ``AttributeError: 'Namespace' object has no attribute 'corpdir'``, before any XML files, database connection, or corpus data are even considered. There is currently no way to invoke this switch successfully from the command line.

Other Switches
==============

Required Switches:

* None (flag currently cannot complete successfully — see Known issue above)

Example Commands
================

Not usable as of this writing — see Known issue above.