---
license: mit
language:
- multilingual
tags:
- zero-shot-classification
- text-classification
- pytorch
metrics:
- accuracy
- f1-score
widget:
- text: We will place an immediate 6-month halt on the finance driven closure of beds
    and wards, and set up an independent audit of needs and facilities.
- text: Разграничение компетенции в сфере социальной политики между федеральными и
    местными властями должно завершиться подписанием специального договора между Центром
    и субъектами федерации о принципах, целях и механизмах социальной политики.
- text: Vse lokalno ali malo širše od lokalnega se skuša prevaliti na državno raven.
- text: Սոցիալ-տնտեսական ասպեկտով ռեժիմի գործունեության արդյունքը եղել է այն, որ 10
    ամենահարուստ ընտանիքների ունեցվածքի շուկայական արժեքը ՀՀ-ում կազմում է ՀՆԱ-ի 54.7
    տոկոսը։
- text: Povedano drugače, na medijski trg ne morejo vplivati oglaševalska sredstva
    podjetij, ki so povsem v zasebni lasti in brez povezav s politiko, ker so, prvič,
    premajhna (rast teh podjetij je bila omejena z rastjo preferenčnih državnih podjetij)
    in drugič, ker njihovo oglaševanje ne more prinašati skritih motivov in agend,
    saj jih ne morejo imeti, ko pa je njihova edina dejavnost tržna in ne politična.
- text: U 2009. I 2010 godini osmišljen je i proveden novi projekt GAZELE temeljem
    kojega je ispladeno 79 potpora u iznosu od 3.657.534 EUR-a.
- text: Toleranţa nu include obligaţia de a tolera intoleranţa sau manifestările antisociale.
- text: edistää sekä yritysten hakeutumista työvoiman luo että työvoiman hakeutumista
    yritysten luo tehokkaalla aluepolitiikalla sekä parantamalla työvoiman ammatillista
    ja alueellista liikkuvuutta.
- text: με εξορθολογισμό της διοίκησης και των αμοιβών της
- text: Från ca 18 barn i början av 1980-talet till 40,4 barn år 2013.
extra_gated_fields:
  Name: text
  Country: country
  Institution: text
  Institution Email: text
  Please specify your academic use case: text
extra_gated_prompt: Our models are intended for academic use only. If you are not
  affiliated with an academic institution, please provide a rationale for using our
  models. Please allow us a few business days to manually review subscriptions.
---
# xlm-roberta-large-manifesto
## Model description
An `xlm-roberta-large` model finetuned on multilingual training data labeled using the [Manifesto Project](https://manifesto-project.wzb.eu/)'s coding scheme. 
This model uses **[Version 2020b (December 23, 2020)](https://manifesto-project.wzb.eu/down/data/2020b/codebooks/codebook_MPDataset_MPDS2020b.pdf)** of the Manifesto Project Dataset Codebook.

## How to use the model

```python
from transformers import AutoTokenizer, pipeline

tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large")
pipe = pipeline(
    model="poltextlab/xlm-roberta-large-manifesto",
    task="text-classification",
    tokenizer=tokenizer,
    use_fast=False,
    token="<your_hf_read_only_token>"
)

text = "We will place an immediate 6-month halt on the finance driven closure of beds and wards, and set up an independent audit of needs and facilities."
pipe(text)
```

### Gated access
Due to the gated access, you must pass the `token` parameter when loading the model. In earlier versions of the Transformers package, you may need to use the `use_auth_token` parameter instead.

## Model performance
The model was evaluated on a test set of 305141 examples, which were split in a stratified manner, where for every label, 20% of all occurences were randomly selected.<br>
Metrics (precision, recall and F1-score are weighted macro averages):

| Precision | Recall | F1-Score | Accuracy | Top3_Acc | Top5_Acc |
|:---------:|:------:|:--------:|:--------:|:--------:|:--------:|
| 0.6495 | 0.6547 | 0.6507 | 0.6547 | 0.8505 | 0.9073 |

Language-specific metrics:
![Model benchmark (language-specific test)](benchmark_model_pooled_low_res.png)

## Debugging and issues
This architecture uses the `sentencepiece` tokenizer. In order to run the model before `transformers==4.27` you need to install it manually.