File size: 4,108 Bytes
f24ea3f 1ba5c03 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 |
---
tags:
- spacy
- token-classification
language:
- da
license: apache-2.0
model-index:
- name: da_dacy_large_DANSK_ner
results:
- task:
name: NER
type: token-classification
metrics:
- name: NER Precision
type: precision
value: 0.8100263852
- name: NER Recall
type: recall
value: 0.8203072812
- name: NER F Score
type: f_score
value: 0.8151344175
---
<a href="https://github.com/centre-for-humanities-computing/Dacy"><img src="https://centre-for-humanities-computing.github.io/DaCy/_static/icon.png" width="175" height="175" align="right" /></a>
# DaCy_large_DANSK_ner
DaCy is a Danish language processing framework with state-of-the-art pipelines as well as functionality for analyzing Danish pipelines.
At the time of publishing this model, also included in DaCy encorporates the only models for fine-grained NER using DANSK dataset - a dataset containing 18 annotation types in the same format as Ontonotes.
Moreover, DaCy's largest pipeline has achieved State-of-the-Art performance on Named entity recognition, part-of-speech tagging and dependency parsing for Danish on the DaNE dataset.
Check out the [DaCy repository](https://github.com/centre-for-humanities-computing/DaCy) for material on how to use DaCy and reproduce the results.
DaCy also contains guides on usage of the package as well as behavioural test for biases and robustness of Danish NLP pipelines.
| Feature | Description |
| --- | --- |
| **Name** | `da_dacy_large_DANSK_ner` |
| **Version** | `0.1.0` |
| **spaCy** | `>=3.5.0,<3.6.0` |
| **Default Pipeline** | `transformer`, `ner` |
| **Components** | `transformer`, `ner` |
| **Vectors** | 0 keys, 0 unique vectors (0 dimensions) |
| **Sources** | DANSK - Danish Annotations for NLP Specific TasKs<br />KennethEnevoldsen/dfm-bert-large-v1-2048bsz-1Msteps (Kenneth Enevoldsen) |
| **License** | `apache-2.0` |
| **Author** | [Centre for Humanities Computing Aarhus](https://chcaa.io/#/) |
### Label Scheme
<details>
<summary>View label scheme (18 labels for 1 components)</summary>
| Component | Labels |
| --- | --- |
| **`ner`** | `CARDINAL`, `DATE`, `EVENT`, `FACILITY`, `GPE`, `LANGUAGE`, `LAW`, `LOCATION`, `MONEY`, `NORP`, `ORDINAL`, `ORGANIZATION`, `PERCENT`, `PERSON`, `PRODUCT`, `QUANTITY`, `TIME`, `WORK OF ART` |
</details>
### Accuracy
| Type | Score |
| --- | --- |
| `ENTS_F` | 81.51 |
| `ENTS_P` | 81.00 |
| `ENTS_R` | 82.03 |
| `TRANSFORMER_LOSS` | 63375.61 |
| `NER_LOSS` | 158164.20 |
### Performance tables
The table below shows the F1, recall and precision of the three DaCy fine-grained models.
| Score | DaCy large | DaCy medium | DaCy small |
|:---:|:---:|:---:|:---:|
| F1 | 0.823 | 0.806 | 0.776 |
| Recall | 0.834 | 0.818 | 0.77 |
| Precision | 0.813 | 0.794 | 0.781 |
The table below shows the F1 of the three DaCy fine-grained models within each named entity type.
| Named-entity type | DaCy large | DaCy medium | DaCy small |
|:---:|:---:|:---:|:---:|
| CARDINAL | 0.874 | 0.781 | 0.887 |
| DATE | 0.846 | 0.859 | 0.867 |
| EVENT | 0.611 | 0.571 | 0.4 |
| FACILITY | 0.545 | 0.533 | 0.468 |
| GPE | 0.893 | 0.838 | 0.794 |
| LANGUAGE | 0.902 | 0.486 | 0.194 |
| LAW | 0.686 | 0.625 | 0.606 |
| LOCATION | 0.633 | 0.737 | 0.581 |
| MONEY | 0.993 | 1 | 0.947 |
| NORP | 0.78 | 0.887 | 0.785 |
| ORDINAL | 0.696 | 0.7 | 0.727 |
| ORGANIZATION | 0.863 | 0.851 | 0.781 |
| PERCENT | 0.923 | 0.96 | 0.96 |
| PERSON | 0.871 | 0.872 | 0.833 |
| PRODUCT | 0.671 | 0.635 | 0.526 |
| QUANTITY | 0.386 | 0.654 | 0.708 |
| TIME | 0.643 | 0.571 | 0.71 |
| WORK OF ART | 0.494 | 0.639 | 0.488 |
The table below shows the F1 of the three DaCy fine-grained models within each domain of texts in DANSK.
| Domain | DaCy large | DaCy medium | DaCy small |
|:---:|:---:|:---:|:---:|
| All domains combined | 0.823 | 0.806 | 0.776 |
| Conversation | 0.796 | 0.718 | 0.82 |
| Dannet | 0.75 | 0.667 | 1 |
| Legal | 0.852 | 0.854 | 0.866 |
| News | 0.841 | 0.759 | 0.86 |
| Social Media | 0.793 | 0.847 | 0.8 |
| Web | 0.826 | 0.802 | 0.756 |
| Wiki and Books | 0.778 | 0.838 | 0.709 |
|