frontend

app.py

@@ -0,0 +1,84 @@
+import time
+
+import streamlit as st
+
+from categories.accuracy import *
+
+
+def response_generator(prompt):
+    source = st.session_state.german
+    acc = accuracy(source, prompt)
+
+    response = "Your response is: " + str(acc["score"]) + "\n"
+
+    if acc["errors"]:
+        response += "Your errors are:\n"
+
+        for error in acc["errors"]:
+            response += f" - {error['message']}\n"
+
+    lines = response.split("\n")
+    for line in lines:
+        for word in line.split():
+            yield word + " "
+            time.sleep(0.05)
+        # After each line, yield a newline character or trigger a line break in Markdown
+        yield "\n"
+
+
+def translation_generator():
+    st.session_state.german = "Danke shoen."
+
+    message = (
+        f"Please translate the following sentence into English:"
+        f" {st.session_state.german}"
+    )
+
+    lines = message.split("\n")
+    for line in lines:
+        for word in line.split():
+            yield word + " "
+            time.sleep(0.05)
+        # After each line, yield a newline character or trigger a line break in Markdown
+        yield "\n"
+
+
+st.title("Translation bot")
+
+# Initialize chat history
+if "messages" not in st.session_state:
+    st.session_state.messages = [
+        {
+            "role": "assistant",
+            "content": (
+                "Hello! I am a translation bot. Please translate the following"
+                " sentence into English: 'Das ist ein Test.'"
+            ),
+        }
+    ]
+    st.session_state.german = "Das ist ein Test."
+
+# Display chat messages from history on app rerun
+for message in st.session_state.messages:
+    with st.chat_message(message["role"]):
+        st.markdown(message["content"])
+
+# Accept user input
+if prompt := st.chat_input("What is up?"):
+    # Add user message to chat history
+    st.session_state.messages.append({"role": "user", "content": prompt})
+    # Display user message in chat message container
+    with st.chat_message("user"):
+        st.markdown(prompt)
+
+    # Display assistant response in chat message container
+    with st.chat_message("assistant"):
+        response = st.write_stream(response_generator(prompt))
+
+    st.session_state.messages.append({"role": "assistant", "content": response})
+
+    with st.chat_message("assistant"):
+        message = st.write_stream(translation_generator())
+
+    st.session_state.messages.append({"role": "assistant", "content": message})
+    # Add assistant response to chat history

categories/accuracy.py

@@ -1,6 +1,7 @@
 import string
 
 import torch
+import numpy as np
 from scipy.spatial.distance import cosine
 from simalign import SentenceAligner
 from transformers import AutoModel, AutoTokenizer
@@ -97,7 +98,9 @@ def __bertscore_to_percentage(similarity: float) -> float:
         int: A score from 0 to 100.
     """
     # Scale the similarity score from [-1, 1] range to [0, 100] (rarely negative)
-    scaled_score = max(((similarity) / 2) * 100, 0)
+    # Logistic function: 100 / (1 + exp(-k * (x - 0.5))), where k controls steepness
+    k = 35  # Steepness parameter - higher values create a sharper transition
+    scaled_score = 100 / (1 + np.exp(-k * (similarity - 0.65)))
     return round(scaled_score, 2)
 
 

requirements.txt

@@ -2,4 +2,5 @@ language_tool_python
 transformers
 torch
 wordfreq
-simalign
+simalign
+streamlit

scorer.ipynb

@@ -81,14 +81,14 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 5,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Fluency Score: 76.61500000000001\n",
+      "Fluency Score: 76.62\n",
       "Accuracy Score: 24.45\n"
      ]
     }