Chapter 4 — Further Reading

Sources are grouped by the book's three evidence tiers (see Chapter 2 and the style guide). Tier 1 is verified/canonical; Tier 2 is real industry material whose exact figures we don't personally verify; Tier 3 is this book's own constructed teaching material. Read the Tier-1 items first — they are the ground truth everything else interprets.

Tier 1 — Verified / canonical

• Google, "Introducing the Knowledge Graph: things, not strings" (Amit Singhal, Google official blog, 2012). The primary source for the entity shift and the "things, not strings" framing, including Google's own launch figures for the graph's scale. Start here.
• Google, "Understanding searches better than ever before" (2019). Google's own explanation of the BERT update, with its worked examples (the "brazil traveler to usa" visa query; "parking on a hill with no curb") showing why word order and function words change meaning.
• Google Search Central — Structured data / Organization and sameAs documentation. The canonical reference for how to declare entities and link them to authoritative references. (Full schema treatment is Chapter 18; this is the primary doc.)
• Google Search Central — "How Google Search works" and the Search Quality Rater Guidelines. Background on how Google understands queries and content; the Rater Guidelines describe how human raters assess whether a page meets a query's meaning and need (E-E-A-T is Chapter 5).
• Google, knowledge panel help documentation ("Get a knowledge panel" / claiming and suggesting changes). The official, honest account of what you can and cannot do about your own panel.
• Deerwester, Dumais, et al., "Indexing by Latent Semantic Analysis" (1990). The original LSI paper — read it to see what the technique actually is, and why "LSI keywords" is a misappropriation of its name.
• Robertson & Zaragoza, "The Probabilistic Relevance Framework: BM25 and Beyond" (2009). The definitive reference on BM25 — where the saturation and length-normalization terms come from, in the original information-retrieval context. Technical, but the source of truth for the formula in §4.4.
• Spärck Jones (1972), "A statistical interpretation of term specificity and its application in retrieval." The origin of inverse document frequency — the "rare terms matter more" idea at the heart of TF-IDF.
• Wikidata (wikidata.org). Not a reading so much as a tool: browse how a real entity's facts and relationships are structured. It is one of the corroborating sources Google draws on.

Tier 2 — Attributed; specifics unverified

• John Mueller (Google Search Advocate), public statements on "LSI keywords" (social media, 2019, widely cited). The clearest on-record debunking from a Google representative — useful precisely because it is a Google voice, though it is a tweet, not documentation.
• Practitioner writing on "entity SEO" and "topical authority" (e.g., established SEO publications and authors such as those at Search Engine Journal, Search Engine Land, and Moz). Treat their frameworks as informed professional models, not confirmed Google mechanics — and watch for the myths (LSI, TF-IDF-tool magic) that recur even in reputable outlets.
• Vendor "topic/keyword clustering" and content-optimization tools (Surfer, Clearscope, MarketMuse, and similar). Genuinely useful for coverage mapping if used to check "did I address the concepts an expert would?" — and misused when treated as "hit these term counts." Read §4.4's myth callout before trusting any "score."

Tier 3 — Illustrative / constructed (this book)

• Rivertown Home Services — the running project; the entity-cluster territory map (Heating / Cooling / Plumbing / Electrical) in this chapter's Strategy File is a constructed teaching example.
• Figures 4.1–4.2 — the knowledge-panel rendering and the "one deep guide, forty queries" report are constructed teaching examples; all numbers are illustrative.

Suggested order

1. Google's "things, not strings" post — the founding statement of the whole shift.
2. Google's BERT explainer — see semantic/natural-language understanding in Google's own words.
3. Google Search Central on structured data / sameAs — the practical entity-linking channel.
4. Skim the Spärck Jones (IDF) idea and the BM25 reference for §4.4's intuition (don't get lost in the algebra — you want the two ideas, saturation and length normalization).
5. Read the original LSI paper's abstract, then Mueller's debunking — the cleanest lesson in evidence over folklore in this book.
6. Finally, browse a real entity on Wikidata to see corroborated facts and relationships in the wild.