Overall, it’s better to use multi-dimensional ontology, not a flat typology:

(a) Locus/topic/operator being stressed: Truth (T), Belief (B), Knowledge (K), Provability (Prov), Obligation (O), Probability (Pr), Utility (U), Reference (Ref), Computation (Comp), Time (Temp), Set/∞.

(b) Mechanism: Self-reference, diagonalization, fixed point, vagueness, aggregation, infinity/size, dominance/expectation, conditioning, causal loop, underspecification/context-shift.

Many paradoxes use the same schema (or mechanism), but vary by degree and topic.

For example, “Liar Paradox” (fixed point/diagonalization):

• “This sentence is false”
• Label Q: “Sentence with label Q is false”, Cards, Yablo’s paradoxes
• X = ¬X
• C = C->A (Curry’s)
• λ ≡ ¬T(⟦λ⟧)
• Gödel (G ≡ ¬Prov(⟦G⟧)
• Chaitin’s Incompleteness,
• Epistemic: κ ≡ ¬K(⟦κ⟧)
• Kripke’s fixed point, Gupta/Belnap (The Revision Theory of Truth)

(c) Modality/logic used: Alethic, epistemic, doxastic, deontic, temporal, provability, paraconsistent, relevant, substructural, many-valued, supervaluationist, etc.

(d) Phenotype of failure: Inconsistency (explosive vs paraconsistent), indeterminacy, triviality, non-computability, undecidability, non-derivability, collapse of distinctions (e.g., Curry -> triviality without T).

(e) Preferred resolution strategy: Hierarchies (Tarski), paraconsistency, contextualism/dynamic semantics, type theory/stratification, supervaluationism, restriction of inference rules (e.g., contraction), proof-theoretic semantics, choice under ambiguity (maximin, causal vs evidential decision theory), etc.

For example,

• Curry = {locus: Prov/Truth/Deontic (depending on variant); mechanism: fixed point + contraction; phenotype: triviality; resolution: restrict contraction or detatchment, or use substructural logic}.
• Liar = {locus: Truth; mechanism: fixed point/diagonalization; phenotype: inconsistency/triviality (classical); resolution: hierarchy, paraconsistency, Kripke fixed points,…}.
• Moore = {locus: Belief & Pragmatics; mechanism: assertion-belief misalignment; phenotype: pragmatic absurdity; resolution: norms of assertion / belief introspection constraints}.
• Sorites = {locus: meaning/vagueness; mechanism: tolerance + induction; phenotype: indeterminacy or inconsistency with classical tolerance; resolution: supervaluationism, degree semantics, epistemicism, contextualism}.
• Fitch = {locus: knowledge; mechanism: modal embedding + knowability; phenotype: collapse to omniscience; resolution: restrict knowability schema, anti-realist moves, proof-theoretic tweaks}.

Additionally, there are various Formalization Degrees: the more formal it becomes the easier it is to transition from “cheap talk” to paradoxes, from paradoxes to theorems, from theorems to theories.

And also, we distinguish different phenomena:

• quasi-paradoxes (e.g. Moore’s as p & ¬B(p), or Liar as X = ¬X in the context of classical/binary logic) - they are interesting, but not really paradoxical, or too obvious to resolve.
• paradoxes (which not that easily resolvable),
• theorems (e.g. Gödel’s Incompleteness Theorems),
• theories (e.g. Kripke’s fixed point, The Revision Theory of Truth by Gupta/Belnap),
• and, simply, interesting phenomena or ideas (Meinong Jungle, for example)