import streamlit as st
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
import pandas as pd
# Interface utilisateur
st.set_page_config(
page_title="Traduction d'une phrase en pictogrammes ARASAAC",
page_icon="📝",
layout="wide"
)
# Charger le modèle et le tokenizer
checkpoint = "Propicto/t2p-t5-large-orfeo"
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForSeq2SeqLM.from_pretrained(checkpoint)
# Lire le lexique
@st.cache_data
def read_lexicon(lexicon):
df = pd.read_csv(lexicon, sep='\t')
df['keyword_no_cat'] = df['lemma'].str.split(' #').str[0].str.strip().str.replace(' ', '_')
return df
lexicon = read_lexicon("lexicon.csv")
# Processus de sortie de la traduction
def process_output_trad(pred):
return pred.split()
def get_id_picto_from_predicted_lemma(df_lexicon, lemma):
id_picto = df_lexicon.loc[df_lexicon['keyword_no_cat'] == lemma, 'id_picto'].tolist()
return (id_picto[0], lemma) if id_picto else (0, lemma)
# Génération du contenu HTML pour afficher les pictogrammes
def generate_html(ids):
html_content = ''
for picto_id, lemma in ids:
if picto_id != 0: # ignore invalid IDs
img_url = f"https://static.arasaac.org/pictograms/{picto_id}/{picto_id}_500.png"
html_content += f'''
{lemma}
'''
html_content += ''
return html_content
st.title("Traduction d'une phrase en pictogrammes ARASAAC")
sentence = st.text_input("Entrez une phrase en français:")
if sentence:
inputs = tokenizer(sentence, return_tensors="pt").input_ids
outputs = model.generate(inputs, max_new_tokens=40, do_sample=True, top_k=30, top_p=0.95)
pred = tokenizer.decode(outputs[0], skip_special_tokens=True)
sentence_to_map = process_output_trad(pred)
pictogram_ids = [get_id_picto_from_predicted_lemma(lexicon, lemma) for lemma in sentence_to_map]
html = generate_html(pictogram_ids)
st.components.v1.html(html, height=800, scrolling=True)