Book recommender: scoring algorithm

Scoring algorithm design for the book-recommender project. Pure logic, no AI, no embeddings.

Per-book scores (0-100 each)

1. Topic match (weight: 30%)

Compare book’s genre/subjects against core interest threads:

• How things connect: systems, ecology, relationships, taxonomy, community, connection
• The craft of language: language, storytelling, words, communication, conversation design
• The deeper layer: soul, humor, intelligence, philosophy, layers

Each keyword match = points. More matches = higher score.

2. Experience prediction (weight: 25%)

Based on audit calibration data: what made finished books work or fail.

• Signals: genre density (academic vs narrative), page count, known-loved author (Pratchett, Murakami)
• Penalize: “textbook feel” genres (Linguistics, Semiotics, Economics) unless mood is “deep focus”
• Bonus: humor, satire, short stories (momentum-friendly)

3. Mood fit (weight: 25%)

User selects mood at runtime:

• Deep focus: boost dense non-fiction, professional, long reads
• Light & curious: boost short fiction, essays, poetry, short stories
• Comfort zone: boost known authors, series continuations, re-reads
• Challenge me: boost unread genres, unfamiliar authors, “should reads”

4. Garden connection (weight: 10%)

• Match book subjects against garden tags
• Bonus if it connects to an active project (tagged project + recent updated date)

5. Freshness (weight: 10%)

• Boost books marked “not on my radar” or “forgot I had it” (ereader invisibility fix)
• Boost recent purchases
• Penalize abandoned books (unless mood is “challenge me”)

Reading mix output

Instead of ranking all books 1-40, output a recommended mix:

• Main read: highest overall score
• Side read: highest score among short/light books (short stories, essays, reference works, children’s books)
• Wildcard: random pick from top 10 that isn’t in the same genre as main/side

Each book card shows: title, author, total score, score breakdown, estimated days to finish, garden connections, and a one-line “why this one”.