Using beam_ner component fails.

[narayanacharya6]

How to reproduce the behaviour

import spacy
nlp = spacy.load("en_core_web_lg")
nlp = spacy.load("en_core_web_lg", exclude=[name for name in nlp.pipe_names])
nlp.add_pipe("beam_ner")
content = "George Washington University Washington DC is a premier university in the US"
doc = nlp(content)

Stacktrace:

Traceback (most recent call last):
  File "/Users/narayan/anaconda3/envs/project-lis/lib/python3.10/site-packages/IPython/core/interactiveshell.py", line 3251, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-2-077b7f4b02c2>", line 6, in <module>
    doc = nlp(content)
  File "/Users/narayan/anaconda3/envs/project-lis/lib/python3.10/site-packages/spacy/language.py", line 1022, in __call__
    error_handler(name, proc, [doc], e)
  File "/Users/narayan/anaconda3/envs/project-lis/lib/python3.10/site-packages/spacy/util.py", line 1617, in raise_error
    raise e
  File "/Users/narayan/anaconda3/envs/project-lis/lib/python3.10/site-packages/spacy/language.py", line 1017, in __call__
    doc = proc(doc, **component_cfg.get(name, {}))  # type: ignore[call-arg]
  File "spacy/pipeline/trainable_pipe.pyx", line 56, in spacy.pipeline.trainable_pipe.TrainablePipe.__call__
  File "/Users/narayan/anaconda3/envs/project-lis/lib/python3.10/site-packages/spacy/util.py", line 1617, in raise_error
    raise e
  File "spacy/pipeline/trainable_pipe.pyx", line 52, in spacy.pipeline.trainable_pipe.TrainablePipe.__call__
  File "spacy/pipeline/transition_parser.pyx", line 252, in spacy.pipeline.transition_parser.Parser.predict
  File "spacy/pipeline/transition_parser.pyx", line 282, in spacy.pipeline.transition_parser.Parser.beam_parse
  File "spacy/pipeline/_parser_internals/_beam_utils.pyx", line 48, in spacy.pipeline._parser_internals._beam_utils.BeamBatch.__init__
  File "thinc/extra/search.pyx", line 13, in thinc.extra.search.Beam.__init__
AssertionError

Is this the right way to use beam_ner?

Your Environment

• spaCy version: 3.2.3
• Platform: macOS-10.16-x86_64-i386-64bit
• Python version: 3.10.0
• Pipelines: en_core_web_lg (3.2.0)

[adrianeboyd]

add_pipe only adds an uninitialized, untrained component, so first you'd need to initialize the component and train it. (Admittedly this error message is not helpful.)

We recommend using the spacy train CLI for this. You can create a config with spacy init config -p beam_ner beam_ner.config. If you want to use the vectors from en_core_web_lg in your pipeline, then:

spacy init config -p beam_ner beam_ner.config -o accuracy

Toy/example NER training data can be found in several projects including:

• https://github.com/explosion/projects/tree/v3/pipelines/ner_demo
• https://github.com/explosion/projects/tree/v3/pipelines/ner_wikiner

A good place to start in the docs for training: https://spacy.io/usage/training#quickstart

beam_ner isn't included in the website config generator, but you can follow the directions starting right afterwards in the same section.