Update pages/6_Feature_Engineering.py

pages/6_Feature_Engineering.py

@@ -195,44 +195,43 @@ if file_type == "One-Hot Vectorization":
     - One-Hot Vectorization is easy to implement
     ''')
     st.subheader(":red[Disadvantages]")
-    st.markdown("<h1 class='title'>Challenges in One-Hot Vectorization</h1>", unsafe_allow_html=True)
 
-    st.markdown("<h2 class='subtitle'>Different Document Lengths</h2>", unsafe_allow_html=True)
+    st.subheader("blue[Different Document Length]"))
     st.markdown(
     "<p class='content'>Every document contains a different number of words. Here, we are not converting the entire document into a vector, but rather each word separately. "
     "This makes it difficult to structure the data into a tabular format. Converting entire documents into vectors, which is addressed by Bag of Words (BOW), solves this issue.</p>",
     unsafe_allow_html=True,
     )
 
-    st.markdown("<h2 class='subtitle'>Sparsity</h2>", unsafe_allow_html=True)
+    st.subheader(":blue[Sparsity]")
     st.markdown(
     "<p class='content'>The vectors created using one-hot encoding tend to be sparse. When data is given to any algorithm, the model may become biased towards zero values, "
     "leading to an issue in machine learning known as overfitting. This problem is primarily addressed in deep learning.</p>",
     unsafe_allow_html=True,
     )
 
-    st.markdown("<h2 class='subtitle'>Curse of Dimensionality</h2>", unsafe_allow_html=True)
+    st.subheader(":blue[Curse of Dimensionality]")
     st.markdown(
     "<p class='content'>As the number of documents increases, the vocabulary size grows, leading to an increase in dimensionality. This negatively impacts machine learning performance "
     "because the dimensionality of vectors is directly dependent on vocabulary size, which grows as more documents are introduced.</p>",
     unsafe_allow_html=True,
     )
 
-    st.markdown("<h2 class='subtitle'>Out of Vocabulary Issue</h2>", unsafe_allow_html=True)
+    st.subheader(":blue[Out of Vocabulary Issue]")
     st.markdown(
     "<p class='content'>One-hot encoding only converts words that were present in the dataset at the time of training. If a new word appears during inference and was not included in the "
     "training dataset, it cannot be converted into a vector, causing a key error. This issue is effectively solved by FastText.</p>",
     unsafe_allow_html=True,
     )
 
-    st.markdown("<h2 class='subtitle'>Inability to Preserve Semantic Meaning</h2>", unsafe_allow_html=True)
+    st.subheader(":blue[Inability to Preserve Semantic Meaning"])
     st.markdown(
     "<p class='content'>When converting text into vector format, the relationships between words should be preserved. Ideally, similar words should be represented by similar vectors, "
     "meaning the distance between their vectors should be small. If this is achieved, the vectorization method successfully preserves semantic meaning.</p>",
     unsafe_allow_html=True,
     )
 
-    st.markdown("<h2 class='subtitle'>Lack of Sequential Information</h2>", unsafe_allow_html=True)
+    st.subheader(":blue[Lack of Sequential Information]")
     st.markdown(
     "<p class='content'>One-hot encoding does not preserve sequential information in text. The order of words, which is crucial in natural language, is completely lost in this encoding method.</p>",
     unsafe_allow_html=True,