README.md

metadata

license: apache-2.0
language:
  - it
  - fr
  - de
  - es
  - en
base_model:
  - google-bert/bert-base-multilingual-cased
pipeline_tag: text-classification
library_name: transformers

🌍 Multilingual Intent Classifier – Language Switching

This model is a fine-tuned multilingual BERT (bert-base-multilingual-cased) for intent classification of language-switching requests.
It recognizes when a user wants to change the conversation language and supports 5 language:

• english
• italian
• german
• spanish
• french

It recognizes even other class of text like:

• other (generic sentences not related to language switching)
• not_allowed (unsupported languages)

📊 Training Data

• ~6,000 training examples
• Short conversational sentences (e.g. "Can we switch to English?", "Vorrei parlare in italiano", "Nein, bitte auf Deutsch"), and pieaces of conversation steps
• Languages covered: English, Italian, German, Spanish, French
• not_allowed and other provide robustness for real-world inputs

---

🚀 Usage with 🤗 Transformers

You can use the model directly with the pipeline API:

from transformers import pipeline

# Replace with the actual model repo
model_name = "software-si/change-language-intent"

classifier = pipeline(
    task="text-classification",
    model=model_name,
    tokenizer=model_name,
    return_all_scores=True
)

texts = [
    "Vorrei parlare in italiano",
    "Can we switch to English?",
    "Nein, bitte auf Deutsch"
]

results = classifier(texts)

for text, res in zip(texts, results):
    print(f"\nInput: {text}")
    for r in res:
        print(f"  {r['label']}: {r['score']:.4f}")