Named Entity Recognition Node

N

AI/ML/ONNX/NLP

Extract named entities (persons, organizations, locations, dates, etc.) from text using ONNX models. Supports BERT, RoBERTa, and other transformer-based NER models with automatic tokenization. Download models from: BERT-base-NER (https://huggingface.co/dslim/bert-base-NER), Multilingual NER (https://huggingface.co/Davlan/bert-base-multilingual-cased-ner-hrl), spaCy NER (https://huggingface.co/spacy). Text longer than the model's window is split into overlapping chunks rather than truncated, so entities are found throughout a long document. Download tokenizer.json and config.json from the same model repository — config.json carries the id2label mapping that names the entity types and the sequence length the model accepts.

onnx_neronnx

Inputs8

Outputs4

Security exposureUnrated

Packageonnx

Ratings

Scores range from 0 to 10. Higher values mean more impact, exposure, or operational weight.

No score metadata has been set for this node yet.

Input Pins

8

Input

Execution

exec_in

Initiate Execution

Model

Struct

model

ONNX NER Model Session

SchemaNodeOnnxSessionNodeOnnxSessionNodeOnnxSession1 fields

session_refstringrequired

Cache ID for Session

Schema enforced

Tokenizer

Struct

tokenizer

HuggingFace tokenizer.json file for BERT/RoBERTa tokenization. Download from the same model repository.

SchemaFlowPathFlowPathFlowPath3 fields

pathstringrequired

store_refstringrequired

cache_store_refstring | null

Schema enforced

Config

Struct

config

HuggingFace config.json of the model. Supplies the id2label mapping that decides which class index means which entity type, and max_position_embeddings, which sets how many tokens fit in one window. Left empty, the node looks for config.json next to the tokenizer. Strongly recommended: label orderings differ between models of the same size, and a wrong one mislabels every entity.

SchemaFlowPathFlowPathFlowPath3 fields

pathstringrequired

store_refstringrequired

cache_store_refstring | null

Schema enforced

Text

String

text

Input text to analyze for named entities

Labels

String Array

labels

Entity label names in model output order (e.g. ['O', 'B-PER', 'I-PER', 'B-ORG', ...]). Overrides the Config pin. If both are empty, the node falls back to the CoNLL-2003 ordering of dslim/bert-base-NER.

Tagging Scheme

Struct

scheme

Tagging scheme: BIO, BIOES, IOB, or BILOU

SchemaTaggingSchemeTaggingSchemeTaggingScheme4 fields

variant 1constvariant

BIO: Begin, Inside, Outside (most common)

const "BIO"

variant 2constvariant

BIOES: Begin, Inside, Outside, End, Single

const "BIOES"

variant 3constvariant

IOB: Inside, Outside, Begin (legacy format)

const "IOB"

variant 4constvariant

BILOU: Begin, Inside, Last, Outside, Unit

const "BILOU"

Default BIO

BIOBIOESIOBBILOU

Threshold

Float

threshold

Minimum confidence threshold for entity extraction (0.0-1.0)

Default 0.5

Range 0 to 1

Output Pins

4

Output

Execution

exec_out

Done

Result

Struct

result

Full NER result with entities and token predictions

SchemaNerResultNerResultNerResult5 fields

entitiesArray<NamedEntity>required

Recognized entities

itemsNamedEntityarray item

A recognized named entity

textstringrequired

The entity text

entity_typestringrequired

Entity type (PER, ORG, LOC, etc.)

start_charinteger:uintrequired

Character start position in original text

format uintmin 0

end_charinteger:uintrequired

Character end position in original text (exclusive)

format uintmin 0

start_tokeninteger:uintrequired

Start token index

format uintmin 0

+2 more fields

tokensArray<TokenPrediction>required

Token-level predictions

itemsTokenPredictionarray item

Token-level NER prediction

tokenstringrequired

Token text (may include ## for wordpiece)

labelstringrequired

Predicted label (raw from model)

confidencenumber:floatrequired

Confidence score

format float

startinteger:uintrequired

Character offset start

format uintmin 0

endinteger:uintrequired

Character offset end

format uintmin 0

textstringrequired

Original input text

windowinteger:uintrequired

Tokens per window the model actually ran on, including special tokens. Below the model's declared limit means the graph refused that limit and the window was walked down.

format uintmin 0

windowsinteger:uintrequired

Number of overlapping windows the text was split across. More than one means the input was longer than a single pass and was chunked, not truncated.

format uintmin 0

Entities

Struct Array

entities

Extracted named entities as array

SchemaNamedEntityNamedEntityNamedEntity7 fields

textstringrequired

The entity text

entity_typestringrequired

Entity type (PER, ORG, LOC, etc.)

start_charinteger:uintrequired

Character start position in original text

format uintmin 0

end_charinteger:uintrequired

Character end position in original text (exclusive)

format uintmin 0

start_tokeninteger:uintrequired

Start token index

format uintmin 0

end_tokeninteger:uintrequired

End token index (exclusive)

format uintmin 0

confidencenumber:floatrequired

Average confidence score

format float

Count

Integer

entity_count

Number of entities found

Node Info

Internal name

onnx_ner

Category

AI/ML/ONNX/NLP

Version

2