README.md

---
language:
  - el
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - feature-extraction
  - sentence-similarity
  - transformers
metrics:
  - accuracy_cosinus
  - accuracy_euclidean
  - accuracy_manhattan
model-index:
  - name: st-greek-media-bert-base-uncased
    results:
      [
        {
          "task": { "name": "STS Benchmark", "type": "sentence-similarity" },
          "metrics":
            [
              { "type": "accuracy_cosinus", "value": 0.9563965089445283 },
              { "type": "accuracy_euclidean", "value": 0.9566394253292384 },
              { "type": "accuracy_manhattan", "value": 0.9565353183072198 },
            ],
          "dataset":
            {
              "name": "all_custom_greek_media_triplets",
              "type": "sentence-pair",
            },
        },
      ]
---

# Greek Media SBERT (uncased)

## Sentence Transformer

This is a [sentence-transformers](https://www.SBERT.net) based on the [Greek Media BERT (uncased)](https://huggingface.co/dimitriz/greek-media-bert-base-uncased) model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search.

<!--- Describe your model here -->

## Usage (Sentence-Transformers)

Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed:

```

pip install -U sentence-transformers

```

Then you can use the model like this:

```python
from sentence_transformers import SentenceTransformer
sentences = ["This is an example sentence", "Each sentence is converted"]

model = SentenceTransformer('dimitriz/st-greek-media-bert-base-uncased')
embeddings = model.encode(sentences)
print(embeddings)
```

## Usage (HuggingFace Transformers)

Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input
through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word
embeddings.

```python
from transformers import AutoTokenizer, AutoModel
import torch


# Mean Pooling - Take attention mask into account for correct averaging
def mean_pooling(model_output, attention_mask):
    token_embeddings = model_output[0]  # First element of model_output contains all token embeddings
    input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float()
    return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9)


# Sentences we want sentence embeddings for
sentences = ['This is an example sentence', 'Each sentence is converted']

# Load model from HuggingFace Hub
tokenizer = AutoTokenizer.from_pretrained('dimitriz/st-greek-media-bert-base-uncased')
model = AutoModel.from_pretrained('dimitriz/st-greek-media-bert-base-uncased')

# Tokenize sentences
encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt')

# Compute token embeddings
with torch.no_grad():
    model_output = model(**encoded_input)

# Perform pooling. In this case, mean pooling.
sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask'])

print("Sentence embeddings:")
print(sentence_embeddings)
```

## Evaluation Results

<!--- Describe how your model was evaluated -->

For an automated evaluation of this model, see the _Sentence Embeddings
Benchmark_: [https://seb.sbert.net](https://seb.sbert.net?model_name=dimitriz/st-greek-media-bert-base-uncased)

## Training

The model was trained on a custom dataset containing triplets from the **combined** Greek 'internet', 'social-media'
and 'press' domains, described in the paper [DACL](https://...).

- The dataset was created by sampling triplets of sentences from the same domain, where the first two sentences are more
  similar than the third one.
- Training objective was to maximize the similarity between the first two sentences and minimize the similarity between
  the first and the third sentence.
- The model was trained for 3 epochs with a batch size of 16 and a maximum sequence length of 512 tokens.
- The model was trained on a single NVIDIA RTX A6000 GPU with 48GB of memory.

The model was trained with the parameters:

**DataLoader**:

`torch.utils.data.dataloader.DataLoader` of length 10807 with parameters:

```
{'batch_size': 16, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'}
```

**Loss**:

`sentence_transformers.losses.TripletLoss.TripletLoss` with parameters:

```
{'distance_metric': 'TripletDistanceMetric.EUCLIDEAN', 'triplet_margin': 5}
```

Parameters of the fit()-Method:

```
{
    "epochs": 3,
    "evaluation_steps": 1000,
    "evaluator": "sentence_transformers.evaluation.TripletEvaluator.TripletEvaluator",
    "max_grad_norm": 1,
    "optimizer_class": "<class 'torch.optim.adamw.AdamW'>",
    "optimizer_params": {
        "lr": 2e-05
    },
    "scheduler": "WarmupLinear",
    "steps_per_epoch": null,
    "warmup_steps": 17290,
    "weight_decay": 0.01
}
```

## Full Model Architecture

```
SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False})
)
```

## Citing & Authors

The model has been officially released with the article "DACL: A Domain-Adapted Contrastive Learning Approach to Low Resource Language Representations for Document Clustering Tasks".
Dimitrios Zaikis, Stylianos Kokkas and Ioannis Vlahavas.
In: Iliadis, L., Maglogiannis, I., Alonso, S., Jayne, C., Pimenidis, E. (eds) Engineering Applications of Neural Networks. EANN 2023. Communications in Computer and Information Science, vol 1826. Springer, Cham".

If you use the model, please cite the following:

```bibtex
@InProceedings{10.1007/978-3-031-34204-2_47,
author="Zaikis, Dimitrios
and Kokkas, Stylianos
and Vlahavas, Ioannis",
editor="Iliadis, Lazaros
and Maglogiannis, Ilias
and Alonso, Serafin
and Jayne, Chrisina
and Pimenidis, Elias",
title="DACL: A Domain-Adapted Contrastive Learning Approach to Low Resource Language Representations for Document Clustering Tasks",
booktitle="Engineering Applications of Neural Networks",
year="2023",
publisher="Springer Nature Switzerland",
address="Cham",
pages="585--598",
isbn="978-3-031-34204-2"
}


```