final app push

.streamlit/config.toml

@@ -1,6 +1,6 @@
 [theme]
 primaryColor="#e76020"
 backgroundColor="#FDFFFC"
-secondaryBackgroundColor="#6e896a"
+secondaryBackgroundColor="#6E896A"
 textColor="#0f0f0d"
 font="monospace"

Home.py

@@ -1,4 +1,5 @@
 import streamlit as st
+import base64
 import boto3
 
 st.set_page_config(page_title='Auto-BG: The Game Concept Generator', layout='wide')
@@ -20,6 +21,7 @@ def application():
     from datetime import date
     from description_generator import input_manager, model_control
     from pathlib import Path
+    import base64
 
     #S3 Bucket
     session = boto3.Session(aws_access_key_id=st.secrets.accesskey, aws_secret_access_key=st.secrets.secretaccesskey)
@@ -358,23 +360,31 @@ def blog():
     """
     Blog describing the Auto-BG project
     """
-    with open("BGG Blog MD.md", 'r') as blog_md:
-        blog_text = blog_md.read() 
-    st.markdown(blog_text)
-    st.sidebar.subheader('Auto-BG: The Board Game Concept Generator')
-    st.sidebar.write("*This application attempts to augment one step, early in that journey, when the seeds of an idea combine and sprout into a holistic concept.\
-                      By interpreting disparate mechanical and descriptive tags to identify a game concept, Auto-BG uses a custom pipeline of GPT3 and T5 models to create a new description and proposed titles for a game that doesn't exist today.\
-                      These descriptions support designers-to-be as alternatives to current concepts, seeds for future concepts, or any user as, hopefully, an entertaining thought experiment.*")
+    st.write('### Auto-BG: The Board Game Concept Generator')
+    st.write("#### Abstract")
+    st.write("*This application augments one step in the board game design process by generating potential full game concepts from a collection of descriptive tags.\
+             Auto-BG uses a custom pipeline of GPT3 and T5 models to create a new description and proposed titles for a game that doesn't exist today.\
+             These concepts support general users and designers-to-be as alternatives to current concepts, seeds for future concepts, or an entertaining thought experiment.*")
+    
+    # Code adapted from "Display and Download PDF in Streamlit: A Blog Use Case" by My Data Talk, https://towardsdatascience.com/display-and-download-pdf-in-streamlit-a-blog-use-case-5fc1ac87d4b1
+    with open('Auto_BG_blog.pdf', "rb") as blog_file:
+        blog_pdf = base64.b64encode(blog_file.read()).decode('utf-8')
+        blog_display = f'<center><iframe src="data:application/pdf;base64,{blog_pdf}" width=800 height=800 type="application/pdf"></iframe><center>'
+
+    st.markdown(blog_display, unsafe_allow_html=True)
+
+
 
 def about_us():
     """
     About us page describing creators of Auto-BG
     """
-    st.title("About Us")
-    st.sidebar.subheader('Creators of Auto-BG')
-    st.sidebar.write('*With a shared love of data science and board games, we came together and created Auto-BG as part of our Capstone project\
-                     for our "Master of Applied Data Science" program at the University of Michigan.\
-                     We hope you enjoy!*')
+
+    st.write('### Creators of Auto-BG')
+    st.write('*With a shared love of data science and board games, we came together and created Auto-BG as a Capstone project\
+             in the "Master of Applied Data Science" program at the University of Michigan.\
+             We hope you enjoy!*')
+    st.write("\n")
 
     # Columns containing information on each of the creators
     col1, col2, col3 = st.columns([1,1,1])
@@ -384,7 +394,9 @@ def about_us():
         st.subheader('Nick Canu')
         st.write("""
         **University of Michigan**\n
-        *MADS (Master of Applied Data Science)*\n
+        ***Master of Applied Data Science, Class of 2023***\n
+        N. Canu trained, evaluated, & implemented pipeline classes text and title generator models for Auto-BG.\n 
+        Their current board game obsession is Obsession by Kayenta Games.
         """)
     
     with col2:
@@ -392,7 +404,9 @@ def about_us():
         st.subheader('Taylor Druhot')
         st.write("""
         **University of Michigan**\n
-        *MADS (Master of Applied Data Science)*\n
+        ***Master of Applied Data Science, Class of 2023***\n
+        T. Druhot scoped and designed the Streamlit application. He also built the UI front end & title generator final module, and integrated modules into live application.\n 
+        Avid Magic the Gathering Limited player and deck/engine building board games.
         """)
 
     with col3:
@@ -400,7 +414,9 @@ def about_us():
         st.subheader('Sebastian Capp')
         st.write("""
         **University of Michigan**\n
-        *MADS (Master of Applied Data Science)*\n
+        ***Master of Applied Data Science, Class of 2023***\n
+        S. Capp contributed research, text generator prototyping, streamlit application modules, and visualizations.\n 
+        Immense fan of Catan, Coup, and any deduction games.
         """)
 
 def feedback():

Persistent_Data/Feedback.csv

@@ -0,0 +1,2 @@
+,feedback
+0,

Tuning Data/final_tuning.csv

@@ -0,0 +1,10 @@
+vals,precision,recall,f1,rouge1,rouge2,rougeL,bleurt
+.4.6,0.839104,0.823205,0.830972,0.286055,0.032162,0.137387,-0.658947
+.5.6,0.830372,0.822095,0.826054,0.271715,0.02905,0.128324,-0.694956
+.7.6,0.828157,0.82092,0.82436,0.260092,0.026194,0.122711,-0.7378
+.4.7,0.844516,0.821091,0.832507,0.260654,0.031788,0.129768,-0.759018
+.5.7,0.836557,0.823989,0.830086,0.282257,0.039041,0.134946,-0.638258
+.7.7,0.831543,0.81901,0.825074,0.245687,0.029726,0.121534,-0.782465
+.4.8,0.834656,0.822824,0.828583,0.26926,0.0295,0.12909,-0.700632
+.5.8,0.839928,0.825867,0.832719,0.280697,0.036367,0.137291,-0.6554
+.7.8,0.828769,0.822225,0.8253,0.260533,0.026237,0.121529,-0.69857

Tuning Data/final_tuning.parquet.gzip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a93c53f3b4e0caaf39a227adaba92200cad20a6577f482b227d60d524056b894
+size 200221

Tuning Data/mem_test.parquet.gzip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6012869d662ba2e95a27a7486863a03491c15fbc205c8c4001b68735c960ea75
+size 176763

requirements.txt

@@ -9,4 +9,3 @@ spacy==3.5.0
 streamlit==1.20.0
 torch==1.13.1
 transformers==4.26.1
-sentencepiece==0.1.97

title_generator.py

@@ -75,22 +75,22 @@ class Title_Generator:
     def candidate_score(self,candidates,ex_check=None):
         
         
-    
-        pat = re.compile("((?:" + "|".join(map(re.escape, candidates[0]+[cand.upper() for cand in candidates[0]])) + "))")
-        desc = re.sub(pat, "__", candidates[1])
+        if ex_check != None:
+            pat = re.compile("((?:" + "|".join(map(re.escape, candidates[0]+[cand.upper() for cand in candidates[0]])) + "|" + "|".join(ex_check) +"))")
+            desc = re.sub(pat, "__", candidates[1])
+        else:
+            pat = re.compile("((?:" + "|".join(map(re.escape, candidates[0]+[cand.upper() for cand in candidates[0]])) + "))")
+            desc = re.sub(pat, "__", candidates[1])
 
         
         if re.search(re.compile(re.escape("__")), desc):
+            reg = re.compile("("+"|".join(ex_check) + ")")
             hold = candidates[0]
             gen_desc = re.sub(re.compile(re.escape("__")),"",desc)
             candidates = self.candidate_generator(gen_desc)
-            next = candidates[0]+hold
+            next = [cand for cand in candidates[0]+hold if not reg.search(cand)]
             candidates = (next, desc)
 
-        reg = re.compile("("+"|".join(ex_check) + ")")
-        step = [cand for cand in candidates[0] if not reg.search(cand)]
-        candidates = (step,candidates[1])
-
         #check for existing games and duplicates
         #transform function from https://stackoverflow.com/questions/42165779/python-how-to-remove-duplicate-valuescase-insensitive-from-a-list-with-same-o
         def transform(L):