Spaces:
Sleeping
Sleeping
Upload 2 files
Browse files
app.py
ADDED
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from haystack.nodes import FARMReader, PreProcessor, PDFToTextConverter, TfidfRetriever
|
3 |
+
from haystack.document_stores import InMemoryDocumentStore
|
4 |
+
from haystack.pipelines import ExtractiveQAPipeline
|
5 |
+
|
6 |
+
document_store = InMemoryDocumentStore()
|
7 |
+
model = "./artifacts/model-afwukuq2:v0/"
|
8 |
+
reader = FARMReader(model_name_or_path=model)
|
9 |
+
preprocessor = PreProcessor(
|
10 |
+
clean_empty_lines=True,
|
11 |
+
clean_whitespace=True,
|
12 |
+
clean_header_footer=True,
|
13 |
+
split_by="word",
|
14 |
+
split_length=100,
|
15 |
+
split_respect_sentence_boundary=True,
|
16 |
+
split_overlap=3
|
17 |
+
)
|
18 |
+
|
19 |
+
|
20 |
+
def print_answers(results):
|
21 |
+
fields = ["answer", "score"] # "context",
|
22 |
+
answers = results["answers"]
|
23 |
+
filtered_answers = []
|
24 |
+
|
25 |
+
for ans in answers:
|
26 |
+
filtered_ans = {
|
27 |
+
field: getattr(ans, field)
|
28 |
+
for field in fields
|
29 |
+
if getattr(ans, field) is not None
|
30 |
+
}
|
31 |
+
filtered_answers.append(filtered_ans)
|
32 |
+
|
33 |
+
return filtered_answers
|
34 |
+
|
35 |
+
|
36 |
+
def pdf_to_document_store(pdf_file):
|
37 |
+
document_store.delete_documents()
|
38 |
+
converter = PDFToTextConverter(
|
39 |
+
remove_numeric_tables=True, valid_languages=["es"])
|
40 |
+
documents = [converter.convert(file_path=pdf_file, meta=None)[0]]
|
41 |
+
preprocessed_docs = preprocessor.process(documents)
|
42 |
+
document_store.write_documents(preprocessed_docs)
|
43 |
+
return None
|
44 |
+
|
45 |
+
|
46 |
+
def predict(question):
|
47 |
+
pdf_to_document_store("data.pdf")
|
48 |
+
retriever = TfidfRetriever(document_store=document_store)
|
49 |
+
pipe = ExtractiveQAPipeline(reader, retriever)
|
50 |
+
result = pipe.run(query=question, params={"Retriever": {
|
51 |
+
"top_k": 5}, "Reader": {"top_k": 3}})
|
52 |
+
answers = print_answers(result)
|
53 |
+
return answers
|
54 |
+
|
55 |
+
|
56 |
+
title = "Chatbot Refugiados"
|
57 |
+
|
58 |
+
iface = gr.Interface(fn=predict,
|
59 |
+
inputs=[gr.inputs.Textbox(lines=3, label='Haz una pregunta')],
|
60 |
+
outputs="text",
|
61 |
+
title=title,
|
62 |
+
theme="huggingface",
|
63 |
+
examples=['Dónde pedir ayuda?', 'qué hacer al llegar a España?']
|
64 |
+
)
|
65 |
+
iface.launch()
|
data.pdf
ADDED
Binary file (54.2 kB). View file
|
|