Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -3,25 +3,29 @@ from modules.extractive import TFIDFSummarizer, TextRankSummarizer, CombinedSumm
|
|
| 3 |
from modules.abstractive import load_summarizers, abstractive_summary
|
| 4 |
from modules.preprocessing import Preprocessor, PDFProcessor
|
| 5 |
from modules.utils import handle_long_text
|
| 6 |
-
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
|
| 7 |
|
| 8 |
# Cargar modelos abstractivos finetuneados
|
| 9 |
summarizers = load_summarizers()
|
| 10 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
# Función principal para generar resúmenes
|
| 12 |
def summarize(input_text, file, summary_type, method, num_sentences, model_name, max_length, num_beams):
|
| 13 |
preprocessor = Preprocessor()
|
| 14 |
|
| 15 |
-
# Procesar archivo si se carga
|
| 16 |
if file is not None:
|
| 17 |
pdf_processor = PDFProcessor()
|
| 18 |
input_text = pdf_processor.pdf_to_text(file.name)
|
| 19 |
|
| 20 |
-
# Validar entrada de texto
|
| 21 |
if not input_text:
|
| 22 |
return "Por favor, ingrese texto o cargue un archivo válido."
|
| 23 |
|
| 24 |
-
# Limpiar texto de entrada
|
| 25 |
cleaned_text = preprocessor.clean_text(input_text)
|
| 26 |
|
| 27 |
if summary_type == "Extractivo":
|
|
@@ -36,7 +40,6 @@ def summarize(input_text, file, summary_type, method, num_sentences, model_name,
|
|
| 36 |
else:
|
| 37 |
return "Método no válido para resumen extractivo."
|
| 38 |
|
| 39 |
-
# Generar resumen extractivo
|
| 40 |
return summarizer.summarize(
|
| 41 |
preprocessor.split_into_sentences(cleaned_text),
|
| 42 |
preprocessor.clean_sentences(preprocessor.split_into_sentences(cleaned_text)),
|
|
@@ -46,38 +49,30 @@ def summarize(input_text, file, summary_type, method, num_sentences, model_name,
|
|
| 46 |
elif summary_type == "Abstractivo":
|
| 47 |
if model_name not in summarizers:
|
| 48 |
return "Modelo no disponible para resumen abstractivo."
|
| 49 |
-
model, tokenizer = summarizers[model_name]
|
| 50 |
return handle_long_text(
|
| 51 |
cleaned_text,
|
| 52 |
-
|
| 53 |
-
|
| 54 |
max_length=max_length,
|
| 55 |
stride=128,
|
| 56 |
)
|
| 57 |
|
| 58 |
-
|
| 59 |
elif summary_type == "Combinado":
|
| 60 |
if model_name not in summarizers:
|
| 61 |
return "Modelo no disponible para resumen abstractivo."
|
| 62 |
-
|
| 63 |
-
# Paso 1: Generar puntos clave con el método extractivo
|
| 64 |
-
extractive_keypoints = TFIDFSummarizer().summarize(
|
| 65 |
preprocessor.split_into_sentences(cleaned_text),
|
| 66 |
preprocessor.clean_sentences(preprocessor.split_into_sentences(cleaned_text)),
|
| 67 |
num_sentences,
|
| 68 |
)
|
| 69 |
-
|
| 70 |
-
# Paso 2: Generar el resumen abstractivo guiado por los puntos clave
|
| 71 |
-
combined_input = f"{cleaned_text}\n\nPuntos clave: {extractive_keypoints}"
|
| 72 |
return handle_long_text(
|
| 73 |
-
|
| 74 |
-
summarizers[model_name][0],
|
| 75 |
-
summarizers[model_name][1],
|
| 76 |
max_length=max_length,
|
| 77 |
stride=128,
|
| 78 |
)
|
| 79 |
|
| 80 |
-
|
| 81 |
return "Seleccione un tipo de resumen válido."
|
| 82 |
|
| 83 |
# Interfaz dinámica
|
|
@@ -86,17 +81,36 @@ with gr.Blocks() as interface:
|
|
| 86 |
|
| 87 |
# Entrada de texto o archivo
|
| 88 |
with gr.Row():
|
| 89 |
-
input_text = gr.Textbox(
|
| 90 |
file = gr.File(label="Subir archivo (PDF, TXT)")
|
| 91 |
|
| 92 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
summary_type = gr.Radio(
|
| 94 |
["Extractivo", "Abstractivo", "Combinado"],
|
| 95 |
label="Tipo de resumen",
|
| 96 |
value="Extractivo",
|
| 97 |
)
|
| 98 |
-
|
| 99 |
-
# Opciones dinámicas
|
| 100 |
method = gr.Radio(
|
| 101 |
["TF-IDF", "TextRank", "BERT", "TF-IDF + TextRank"],
|
| 102 |
label="Método Extractivo",
|
|
@@ -117,7 +131,6 @@ with gr.Blocks() as interface:
|
|
| 117 |
1, 10, value=4, step=1, label="Número de haces (Abstractivo)", visible=False
|
| 118 |
)
|
| 119 |
|
| 120 |
-
# Actualización dinámica de opciones
|
| 121 |
def update_options(summary_type):
|
| 122 |
if summary_type == "Extractivo":
|
| 123 |
return (
|
|
@@ -135,16 +148,14 @@ with gr.Blocks() as interface:
|
|
| 135 |
gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), gr.update(visible=False),
|
| 136 |
gr.update(visible=False))
|
| 137 |
|
| 138 |
-
|
| 139 |
summary_type.change(
|
| 140 |
update_options,
|
| 141 |
inputs=[summary_type],
|
| 142 |
outputs=[method, num_sentences, model_name, max_length, num_beams],
|
| 143 |
)
|
| 144 |
|
| 145 |
-
# Botón para generar resumen
|
| 146 |
summarize_button = gr.Button("Generar Resumen")
|
| 147 |
-
output = gr.Textbox(lines=10, label="Resumen generado", interactive=
|
| 148 |
copy_button = gr.Button("Copiar Resumen")
|
| 149 |
|
| 150 |
summarize_button.click(
|
|
@@ -153,7 +164,6 @@ with gr.Blocks() as interface:
|
|
| 153 |
outputs=output,
|
| 154 |
)
|
| 155 |
|
| 156 |
-
# Funcionalidad para copiar resumen
|
| 157 |
def copy_summary(summary):
|
| 158 |
return summary
|
| 159 |
|
|
|
|
| 3 |
from modules.abstractive import load_summarizers, abstractive_summary
|
| 4 |
from modules.preprocessing import Preprocessor, PDFProcessor
|
| 5 |
from modules.utils import handle_long_text
|
|
|
|
| 6 |
|
| 7 |
# Cargar modelos abstractivos finetuneados
|
| 8 |
summarizers = load_summarizers()
|
| 9 |
|
| 10 |
+
# Función para procesar el archivo cargado
|
| 11 |
+
def process_file(file):
|
| 12 |
+
if file is not None:
|
| 13 |
+
pdf_processor = PDFProcessor()
|
| 14 |
+
input_text = pdf_processor.pdf_to_text(file.name)
|
| 15 |
+
return input_text
|
| 16 |
+
return "Por favor, cargue un archivo válido."
|
| 17 |
+
|
| 18 |
# Función principal para generar resúmenes
|
| 19 |
def summarize(input_text, file, summary_type, method, num_sentences, model_name, max_length, num_beams):
|
| 20 |
preprocessor = Preprocessor()
|
| 21 |
|
|
|
|
| 22 |
if file is not None:
|
| 23 |
pdf_processor = PDFProcessor()
|
| 24 |
input_text = pdf_processor.pdf_to_text(file.name)
|
| 25 |
|
|
|
|
| 26 |
if not input_text:
|
| 27 |
return "Por favor, ingrese texto o cargue un archivo válido."
|
| 28 |
|
|
|
|
| 29 |
cleaned_text = preprocessor.clean_text(input_text)
|
| 30 |
|
| 31 |
if summary_type == "Extractivo":
|
|
|
|
| 40 |
else:
|
| 41 |
return "Método no válido para resumen extractivo."
|
| 42 |
|
|
|
|
| 43 |
return summarizer.summarize(
|
| 44 |
preprocessor.split_into_sentences(cleaned_text),
|
| 45 |
preprocessor.clean_sentences(preprocessor.split_into_sentences(cleaned_text)),
|
|
|
|
| 49 |
elif summary_type == "Abstractivo":
|
| 50 |
if model_name not in summarizers:
|
| 51 |
return "Modelo no disponible para resumen abstractivo."
|
|
|
|
| 52 |
return handle_long_text(
|
| 53 |
cleaned_text,
|
| 54 |
+
summarizers[model_name][0],
|
| 55 |
+
summarizers[model_name][1],
|
| 56 |
max_length=max_length,
|
| 57 |
stride=128,
|
| 58 |
)
|
| 59 |
|
|
|
|
| 60 |
elif summary_type == "Combinado":
|
| 61 |
if model_name not in summarizers:
|
| 62 |
return "Modelo no disponible para resumen abstractivo."
|
| 63 |
+
extractive_summary = TFIDFSummarizer().summarize(
|
|
|
|
|
|
|
| 64 |
preprocessor.split_into_sentences(cleaned_text),
|
| 65 |
preprocessor.clean_sentences(preprocessor.split_into_sentences(cleaned_text)),
|
| 66 |
num_sentences,
|
| 67 |
)
|
|
|
|
|
|
|
|
|
|
| 68 |
return handle_long_text(
|
| 69 |
+
extractive_summary,
|
| 70 |
+
summarizers[model_name][0],
|
| 71 |
+
summarizers[model_name][1],
|
| 72 |
max_length=max_length,
|
| 73 |
stride=128,
|
| 74 |
)
|
| 75 |
|
|
|
|
| 76 |
return "Seleccione un tipo de resumen válido."
|
| 77 |
|
| 78 |
# Interfaz dinámica
|
|
|
|
| 81 |
|
| 82 |
# Entrada de texto o archivo
|
| 83 |
with gr.Row():
|
| 84 |
+
input_text = gr.Textbox(max_lines=9, label="Ingrese texto", interactive=True)
|
| 85 |
file = gr.File(label="Subir archivo (PDF, TXT)")
|
| 86 |
|
| 87 |
+
# Nuevo botón para cargar el archivo, inicialmente invisible
|
| 88 |
+
load_file_button = gr.Button("Cargar Archivo", visible=False)
|
| 89 |
+
|
| 90 |
+
# Acción del botón: procesar el archivo y colocar el texto en la caja de texto
|
| 91 |
+
load_file_button.click(
|
| 92 |
+
process_file,
|
| 93 |
+
inputs=[file],
|
| 94 |
+
outputs=[input_text],
|
| 95 |
+
)
|
| 96 |
+
|
| 97 |
+
# Mostrar el botón solo cuando se suba un archivo
|
| 98 |
+
def toggle_load_button(file):
|
| 99 |
+
return gr.update(visible=file is not None)
|
| 100 |
+
|
| 101 |
+
file.change(
|
| 102 |
+
toggle_load_button,
|
| 103 |
+
inputs=[file],
|
| 104 |
+
outputs=[load_file_button],
|
| 105 |
+
)
|
| 106 |
+
|
| 107 |
+
# Selección de tipo de resumen y opciones dinámicas
|
| 108 |
summary_type = gr.Radio(
|
| 109 |
["Extractivo", "Abstractivo", "Combinado"],
|
| 110 |
label="Tipo de resumen",
|
| 111 |
value="Extractivo",
|
| 112 |
)
|
| 113 |
+
|
|
|
|
| 114 |
method = gr.Radio(
|
| 115 |
["TF-IDF", "TextRank", "BERT", "TF-IDF + TextRank"],
|
| 116 |
label="Método Extractivo",
|
|
|
|
| 131 |
1, 10, value=4, step=1, label="Número de haces (Abstractivo)", visible=False
|
| 132 |
)
|
| 133 |
|
|
|
|
| 134 |
def update_options(summary_type):
|
| 135 |
if summary_type == "Extractivo":
|
| 136 |
return (
|
|
|
|
| 148 |
gr.update(visible=False), gr.update(visible=False), gr.update(visible=False), gr.update(visible=False),
|
| 149 |
gr.update(visible=False))
|
| 150 |
|
|
|
|
| 151 |
summary_type.change(
|
| 152 |
update_options,
|
| 153 |
inputs=[summary_type],
|
| 154 |
outputs=[method, num_sentences, model_name, max_length, num_beams],
|
| 155 |
)
|
| 156 |
|
|
|
|
| 157 |
summarize_button = gr.Button("Generar Resumen")
|
| 158 |
+
output = gr.Textbox(lines=10, label="Resumen generado", interactive=False)
|
| 159 |
copy_button = gr.Button("Copiar Resumen")
|
| 160 |
|
| 161 |
summarize_button.click(
|
|
|
|
| 164 |
outputs=output,
|
| 165 |
)
|
| 166 |
|
|
|
|
| 167 |
def copy_summary(summary):
|
| 168 |
return summary
|
| 169 |
|