feat: Renumber lectures for X-hour integration (weeks 3-10)

- Week 3: lectures 7,8 → 9,11; added new X-hour lecture 10
- Week 4: lectures 9-11 → 12-14
- Week 5: lectures 12-14 → 15-17
- Week 6: lectures 15-17 → 18-20
- Week 7: lectures 18-20 → 21-23
- Week 9: lectures 21-23 → 24-26
- Week 10: lecture 24 → 27

Updated all internal lecture number references in titles.
New lecture structure: 27 lectures total (4 per week for weeks 1-3).

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <[email protected]>

Author: jeremymanning

slides/week10/lecture24.md slides/week10/lecture27.mdslides/week10/lecture24.md renamed to slides/week10/lecture27.md

@@ -8,7 +8,7 @@ footer: ''
 
 <!-- _class: lead -->
 
-# Lecture 24: Final Project Work Session
+# Lecture 27: Final Project Work Session
 ## Models of Language and Conversation 🤖
 
 **PSYC 51.07: Models of Language and Communication**

slides/week3/lecture10.md

@@ -0,0 +1,382 @@
+---
+marp: true
+theme: cdl-theme
+paginate: true
+header: 'PSYC 51.07: Models of Language and Communication'
+footer: ''
+---
+
+<!-- _class: lead -->
+
+# Lecture 10: X-Hour Embeddings Workshop
+## Week 3: Hands-On Dimensionality Reduction and Word Vectors
+
+**PSYC 51.07: Models of Language and Communication**
+
+---
+
+# Learning Objectives
+
+By the end of this session, you will:
+
+1. Implement classic dimensionality reduction (LSA, LDA)
+2. Train and analyze Word2Vec embeddings
+3. Visualize high-dimensional embeddings using UMAP
+4. Compare different embedding methods on real data
+5. Understand semantic relationships captured by embeddings
+
+**Workshop format:** Hands-on coding with the 20 Newsgroups dataset
+
+---
+
+# Workshop Overview
+
+**Today's Agenda:**
+
+1. **Part 1:** Why embeddings? From sparse to dense representations
+2. **Part 2:** LSA - Latent Semantic Analysis with SVD
+3. **Part 3:** LDA - Latent Dirichlet Allocation for topic modeling
+4. **Part 4:** Word2Vec - Neural word embeddings
+5. **Part 5:** Visualizing embeddings with UMAP
+6. **Part 6:** Comparing methods and document classification
+
+**Companion notebook:** `xhour_embeddings_demo.ipynb`
+
+---
+
+# Part 1: Why Embeddings?
+
+**The problem with sparse representations:**
+
+<div class="columns">
+<div class="column">
+
+**Last week (BoW, TF-IDF):**
+- High dimensional (vocab size)
+- Sparse (mostly zeros)
+- No semantic similarity
+- "dog" and "puppy" are orthogonal
+
+</div>
+<div class="column">
+
+**Embeddings:**
+- Low dimensional (50-300 dims)
+- Dense (all non-zero)
+- Similar words cluster together
+- "dog" and "puppy" are close!
+
+</div>
+</div>
+
+---
+
+# The Magic of Word Vectors
+
+**Famous example:** king - man + woman = queen
+
+<div class="callout info">
+<div class="callout-title">Vector Arithmetic</div>
+
+Word embeddings capture semantic relationships as directions in space:
+- Gender direction: woman - man
+- Royalty direction: king - queen
+- Pluralization: words - word
+
+</div>
+
+**Key insight:** Meaning encoded as geometry!
+
+---
+
+# Part 2: Latent Semantic Analysis (LSA)
+
+**Using SVD to find latent topics:**
+
+$$X \approx U_k \Sigma_k V_k^T$$
+
+<div class="columns">
+<div class="column">
+
+**Algorithm:**
+1. Build TF-IDF matrix $X$
+2. Apply Singular Value Decomposition
+3. Keep top $k$ dimensions
+4. Use $U_k$ as word embeddings
+
+</div>
+<div class="column">
+
+**Interpretation:**
+- $U$: word-topic associations
+- $\Sigma$: topic strengths
+- $V^T$: doc-topic associations
+
+</div>
+</div>
+
+---
+
+# LSA in Code
+
+```python
+from sklearn.decomposition import TruncatedSVD
+from sklearn.feature_extraction.text import TfidfVectorizer
+
+# Build TF-IDF matrix
+tfidf = TfidfVectorizer(max_features=5000, stop_words='english')
+tfidf_matrix = tfidf.fit_transform(documents)
+
+# Apply LSA
+lsa = TruncatedSVD(n_components=100, random_state=42)
+doc_embeddings = lsa.fit_transform(tfidf_matrix)
+word_embeddings = lsa.components_.T
+
+print(f"Explained variance: {lsa.explained_variance_ratio_.sum():.2%}")
+```
+
+**Try it:** Find similar words using cosine similarity!
+
+---
+
+# Part 3: LDA for Topic Modeling
+
+**A probabilistic approach:**
+
+<div class="callout tip">
+<div class="callout-title">Generative Story</div>
+
+LDA imagines documents are created by:
+1. Choosing a mixture of topics
+2. For each word, picking a topic
+3. Sampling a word from that topic
+
+</div>
+
+**Key difference from LSA:**
+- Probabilistic interpretation
+- Non-negative weights
+- More interpretable topics
+
+---
+
+# LDA Example Output
+
+```python
+Topic 0: hockey, game, team, player, season, nhl, play
+Topic 1: space, nasa, launch, orbit, shuttle, satellite
+Topic 2: computer, software, program, file, windows, system
+Topic 3: medical, doctor, patient, disease, health, treatment
+Topic 4: government, president, congress, law, political
+```
+
+<div class="callout info">
+
+Each document is a **mixture** of topics:
+Document #42: 60% Space + 25% Computer + 15% Other
+
+</div>
+
+---
+
+# Part 4: Word2Vec
+
+**Learning embeddings from context:**
+
+<div class="columns">
+<div class="column">
+
+**Skip-gram:**
+Given target word, predict context
+
+"The **cat** sat on mat"
+- cat → the, sat, on
+
+**CBOW:**
+Given context, predict target
+
+the, sat, on → **cat**
+
+</div>
+<div class="column">
+
+```python
+from gensim.models import Word2Vec
+
+model = Word2Vec(
+    sentences=tokenized_docs,
+    vector_size=100,
+    window=5,
+    min_count=5,
+    sg=1  # Skip-gram
+)
+```
+
+</div>
+</div>
+
+---
+
+# Word2Vec: Semantic Similarity
+
+```python
+# Find similar words
+model.wv.most_similar('computer', topn=5)
+# [('software', 0.82), ('program', 0.79), ('system', 0.75), ...]
+
+# Word analogies
+model.wv.most_similar(
+    positive=['woman', 'king'],
+    negative=['man']
+)
+# [('queen', 0.71), ...]
+```
+
+<div class="callout warning">
+<div class="callout-title">Hands-on Exercise</div>
+
+Try creating your own word analogies! What works? What fails?
+
+</div>
+
+---
+
+# Part 5: Visualizing with UMAP
+
+**Projecting 100D → 2D:**
+
+```python
+import umap
+
+reducer = umap.UMAP(
+    n_neighbors=15,
+    min_dist=0.1,
+    metric='cosine'
+)
+
+embeddings_2d = reducer.fit_transform(word_vectors)
+```
+
+**UMAP advantages:**
+- Faster than t-SNE
+- Preserves global structure
+- Better cluster separation
+
+---
+
+# What You Should See
+
+When you visualize embeddings:
+
+<div class="columns">
+<div class="column">
+
+**Sports cluster:**
+- hockey, baseball, player, team, game
+
+**Space cluster:**
+- nasa, shuttle, orbit, launch, space
+
+</div>
+<div class="column">
+
+**Tech cluster:**
+- computer, software, program, windows
+
+**Medical cluster:**
+- doctor, patient, hospital, treatment
+
+</div>
+</div>
+
+<div class="callout tip">
+
+Related words should cluster together even though we never told the model they were related!
+
+</div>
+
+---
+
+# Part 6: Comparing Methods
+
+| Method | Speed | Interpretability | Quality | Data Needed |
+|--------|-------|------------------|---------|-------------|
+| LSA    | Fast  | Medium           | Medium  | Small-Medium |
+| LDA    | Medium | High            | Medium  | Medium |
+| Word2Vec | Medium | Low           | High    | Large |
+
+**Recommendations:**
+- **Quick exploration:** LSA
+- **Interpretable topics:** LDA
+- **Best semantic quality:** Word2Vec
+
+---
+
+# Document Classification with Embeddings
+
+**Using embeddings as features:**
+
+```python
+def document_vector(doc, model):
+    """Average word vectors for document."""
+    tokens = preprocess(doc)
+    vectors = [model.wv[w] for w in tokens if w in model.wv]
+    return np.mean(vectors, axis=0) if vectors else np.zeros(100)
+
+# Train classifier
+X_train = [document_vector(doc, w2v) for doc in train_docs]
+clf = LogisticRegression()
+clf.fit(X_train, y_train)
+```
+
+**Compare to TF-IDF baseline!**
+
+---
+
+# Key Takeaways
+
+1. **Embeddings capture semantic meaning** - similar words have similar vectors
+
+2. **Different methods, different strengths:**
+   - LSA: Fast, linear, interpretable
+   - LDA: Probabilistic, topic-focused
+   - Word2Vec: Neural, best for similarity
+
+3. **Visualization reveals structure** - UMAP shows semantic clusters
+
+4. **Limitations:**
+   - Static (one vector per word, no context)
+   - Requires substantial data
+   - Can encode biases
+
+**Next week:** Contextual embeddings (BERT, GPT)!
+
+---
+
+# Discussion Questions
+
+1. **Why does vector arithmetic work?** What does "king - man + woman" really mean geometrically?
+
+2. **Bias in embeddings:** If Word2Vec learns from news articles, what biases might it capture?
+
+3. **Window size matters:** What happens with window=2 vs window=10?
+
+4. **Out-of-vocabulary problem:** How do you handle words not in your vocabulary?
+
+5. **When to use what:** For a sentiment analysis task, would you choose LSA, LDA, or Word2Vec?
+
+---
+
+# Next Steps
+
+**For Assignment 2:**
+- Use embeddings to improve your classifier
+- Compare at least 2 embedding methods
+- Visualize your embeddings
+
+**Coming up in Lecture 11:**
+- Modern neural word embeddings
+- GloVe and FastText
+- Subword tokenization
+
+**Office hours:** Available if you need help with the notebook!