Hugging Face's logo

SlavicNLP
/
slavicner-linking-single-out-large

Text2Text Generation
Transformers
Safetensors
mt5
entity linking
Inference Endpoints
Model card Files Community
slavicner-linking-single-out-large
File size: 4,038 Bytes 
0bef028
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
eea8663
 
 
 
 
 
 
 
 
 
 
 
0bef028
 
 
 
 
 
 
 
1ccdb11
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
0bef028
 
 
 
 
 
 
 
 
 
2431ab3
 
0bef028
 
 
 
 
2431ab3
d4d61e3
 
 
 
1ccdb11
d4d61e3
65b8a48
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d4d61e3
eea8663
 
 
 
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
 
---
language:
- multilingual
- pl
- ru
- uk
- bg
- cs
- sl
datasets:
- SlavicNER
license: apache-2.0
library_name: transformers
pipeline_tag: text2text-generation
tags:
- entity linking
widget:
  - text: pl:Polsce
    example_title: Polish
  - text: cs:Velké Británii
    example_title: Czech
  - text: bg:българите
    example_title: Bulgarian
  - text: ru:Великобританию
    example_title: Russian
  - text: sl:evropske komisije
    example_title: Slovene
  - text: uk:Європейського агентства лікарських засобів
    example_title: Ukrainian
---

# Model description

This is a baseline model for named entity **lemmatization** trained on the single-out topic split of the 
[SlavicNER corpus](https://github.com/SlavicNLP/SlavicNER).


# Resources and Technical Documentation

- Paper: [Cross-lingual Named Entity Corpus for Slavic Languages](https://arxiv.org/pdf/2404.00482), to appear in LREC-COLING 2024.
- Annotation guidelines: https://arxiv.org/pdf/2404.00482
- SlavicNER Corpus: https://github.com/SlavicNLP/SlavicNER


# Evaluation

| **Language** | **Seq2seq** | **Support** |
|:------------:|:-----------:|-----------------:|
| PL           | 75.13       | 2 549            |
| CS           | 77.92       | 1 137            |
| RU           | 67.56       | 18 018           |
| BG           | 63.60       | 6 085            |
| SL           | 76.81       | 7 082            |
| UK           | 58.94       | 3 085            |
| All          | 68.75       | 37 956           |


# Usage

You can use this model directly with a pipeline for text2text generation:

```python
from transformers import pipeline

model_name = "SlavicNLP/slavicner-linking-single-out-large"
pipe = pipeline("text2text-generation", model_name)

texts = ["pl:Polsce", "cs:Velké Británii", "bg:българите", "ru:Великобританию",
         "sl:evropske komisije", "uk:Європейського агентства лікарських засобів"]

outputs = pipe(texts)

ids = [o['generated_text'] for o in outputs]
print(ids)
# ['GPE-Poland', 'GPE-Great-Britain', 'GPE-Bulgaria', 'GPE-Great-Britain',
#  'ORG-European-Commission', 'ORG-EMA-European-Medicines-Agency']
```


# Citation

```latex
@inproceedings{piskorski-etal-2024-cross-lingual,
    title = "Cross-lingual Named Entity Corpus for {S}lavic Languages",
    author = "Piskorski, Jakub  and
      Marci{\'n}czuk, Micha{\l}  and
      Yangarber, Roman",
    editor = "Calzolari, Nicoletta  and
      Kan, Min-Yen  and
      Hoste, Veronique  and
      Lenci, Alessandro  and
      Sakti, Sakriani  and
      Xue, Nianwen",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
    month = may,
    year = "2024",
    address = "Torino, Italy",
    publisher = "ELRA and ICCL",
    url = "https://aclanthology.org/2024.lrec-main.369",
    pages = "4143--4157",
    abstract = "This paper presents a corpus manually annotated with named entities for six Slavic languages {---} Bulgarian, Czech, Polish, Slovenian, Russian,
                and Ukrainian. This work is the result of a series of shared tasks, conducted in 2017{--}2023 as a part of the Workshops on Slavic Natural
                Language Processing. The corpus consists of 5,017 documents on seven topics. The documents are annotated with five classes of named entities.
                Each entity is described by a category, a lemma, and a unique cross-lingual identifier. We provide two train-tune dataset splits
                {---} single topic out and cross topics. For each split, we set benchmarks using a transformer-based neural network architecture
                with the pre-trained multilingual models {---} XLM-RoBERTa-large for named entity mention recognition and categorization,
                and mT5-large for named entity lemmatization and linking.",
}
```

# Contact

Michał Marcińczuk ([email protected])