Upload app.py

app.py

@@ -2,18 +2,23 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 import urllib.request
 import xml.etree.ElementTree as ET
+from transformers import pipeline
 
 # HuggingFace Inference Client
 #client = InferenceClient("meta-llama/Llama-3.3-70B-Instruct")
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
+# Hugging Face Pipeline für Named Entity Recognition (NER)
+nlp = pipeline("ner", model="dslim/bert-base-NER")
 
-# Funktion, um die Eingabe zu bereinigen und einen prägnanten Query zu erstellen
+# Funktion zur Extraktion von Keywords ohne Füllwörter
 def generate_query(input_text):
-    stopwords = ["welche", "gibt", "es", "zum", "thema", "studien", "über", "zu", "dem"]
-    words = input_text.lower().split()
-    query = " ".join([word for word in words if word not in stopwords])
-    return query.strip()
+    entities = nlp(input_text)
+    keywords = [entity['word'] for entity in entities if entity['entity_group'] in ['MISC', 'ORG', 'LOC', 'PER']]
+    return " ".join(keywords).strip()
+
+
+
 
 # Funktion, um relevante Studien von arXiv zu suchen
 def fetch_arxiv_summary(query, sort_by="relevance", sort_order="descending", max_results=20):
@@ -26,13 +31,14 @@ def fetch_arxiv_summary(query, sort_by="relevance", sort_order="descending", max
         summaries = []
         for entry in root.findall(".//{http://www.w3.org/2005/Atom}entry"):
             title = entry.find("{http://www.w3.org/2005/Atom}title")
-            link = entry.find("{http://www.w3.org/2005/Atom}link")
+            link_element = entry.find("{http://www.w3.org/2005/Atom}link[@rel='alternate']")
             summary = entry.find("{http://www.w3.org/2005/Atom}summary")
-            if summary is not None and title is not None and link is not None:
-                summaries.append(f"Titel: {title.text.strip()}\nLink: {link.text.strip()}\nZusammenfassung: {summary.text.strip()}")
+            link = link_element.attrib.get("href") if link_element is not None else "Kein Link verfügbar"
+            if summary is not None and title is not None:
+                summaries.append(f"Titel: {title.text.strip()}\nLink: {link}\nZusammenfassung: {summary.text.strip()}")
         return summaries if summaries else ["Keine relevanten Studien gefunden."]
     except Exception as e:
-        return [f"Fehler beim Abrufen der Studie: {str(e),str(url)}"]
+        return [f"Fehler beim Abrufen der Studie: {str(e)}"]
 
 # Chatbot-Logik mit arXiv-Integration
 def respond(
@@ -59,7 +65,7 @@ def respond(
         if val[1]:
             messages.append({"role": "assistant", "content": val[1]})
 
-    messages.append({"role": "user", "content": f"{message}\nStudies:\n{study_info}"})
+    messages.append({"role": "user", "content": f"{message}\nUse this Kontext:\n{study_info}"})
 
     # Antwort vom Modell generieren
     response = ""