|
2 | 2 |
|
3 | 3 | References supporting the pragmatics layer architecture. |
4 | 4 |
|
| 5 | +## The Gap We're Filling |
| 6 | + |
| 7 | +> "Syntactic tests ask 'does this data obey the formal rules?', while pragmatic tests ask 'is this data actually good enough for this specific use and user?'" |
| 8 | +> — Semiotic DQ Thesis (2022) |
| 9 | +
|
| 10 | +Existing tools cover syntax and semantics. **No standard tools exist for pragmatics.** |
| 11 | + |
5 | 12 | ## Core Framework Papers |
6 | 13 |
|
7 | 14 | ### Semiotic Principles for Metadata Auditing |
8 | 15 | - **Source:** [research.amanote](https://research.amanote.com/publication/f5oI3HMBKQvf0BhivObD/semiotic-principles-for-metadata-auditing-and-evaluation) |
9 | | -- **Relevance:** Concrete auditing framework (syntagm, sign-functions, corpus boundaries) - validates our thread traversal approach |
| 16 | +- **Validates:** Thread traversal as "syntagmatic rules over records" |
10 | 17 |
|
11 | 18 | ### Semiotic DQ for Behavioral Data (2022 Thesis) |
12 | 19 | - **Source:** [diva-portal.org](https://www.diva-portal.org/smash/get/diva2:1737820/FULLTEXT01.pdf) |
13 | | -- **Relevance:** Operationalizes pragmatic indicators (task adequacy, interpretability, context completeness) - validates our latitude concept |
| 20 | +- **Validates:** Latitude concept maps to "unusable / usable with caveats / fit-for-purpose" |
14 | 21 |
|
15 | 22 | ### DataKitchen: Syntax-Semantics-Pragmatics Gap |
16 | 23 | - **Source:** [datakitchen.io](https://datakitchen.io/the-syntax-semantics-and-pragmatics-gap-in-data-quality-validate-testing/) |
17 | | -- **Relevance:** Industry recognition that existing tools cover syntax/semantics but NOT pragmatics - validates our gap analysis |
| 24 | +- **Validates:** Industry recognition that existing tools miss pragmatics |
18 | 25 |
|
19 | 26 | ### Semiotics in Scientific Data Quality |
20 | 27 | - **Source:** [honghuang.myweb.usf.edu](http://honghuang.myweb.usf.edu/pub2/Huang_JIS.pdf) |
21 | | -- **Relevance:** Sign-relations among data, models, interpretations - validates context-as-signs approach |
| 28 | +- **Validates:** Context-as-signs approach |
| 29 | + |
| 30 | +## Architecture Validation |
22 | 31 |
|
23 | | -## Key Validations |
| 32 | +| Our Concept | Literature Support | |
| 33 | +|-------------|-------------------| |
| 34 | +| Pragmatics layer | "fitness-for-use from user/decision perspective" | |
| 35 | +| Latitude levels | "minimum viable quality thresholds per use" | |
| 36 | +| Thread traversal | "syntagmatic rules" + "corpus boundaries" | |
| 37 | +| Pack bundles | "metadata catalog with intended use, known unsuitable uses" | |
| 38 | +| LLM handles syntax/semantics | Schema validators + OWL reasoners exist; pragmatics doesn't | |
24 | 39 |
|
25 | | -| Our Architecture | Literature Support | |
26 | | -|------------------|-------------------| |
27 | | -| Pragmatics = fitness-for-use | Semiotic DQ thesis: "pragmatic tests ask 'is this data actually good enough for this specific use and user?'" | |
28 | | -| Latitude levels (none→full) | Maps to pragmatic thresholds: "unusable," "usable with caveats," "fit-for-purpose" | |
29 | | -| Thread traversal | Auditing framework's "syntagmatic rules over records" | |
30 | | -| Pack as domain bundle | "metadata catalog extended with intended use, known unsuitable uses" | |
31 | | -| LLM handles syntax/semantics | DataKitchen: schema validators + Great Expectations = syntactic; OWL/reasoners = semantic | |
| 40 | +## Existing Toolchains (What We Don't Build) |
32 | 41 |
|
33 | | -## Toolchain Mapping |
| 42 | +**Syntactic:** Great Expectations, dbt tests, JSON Schema, SDMX validators |
| 43 | +**Semantic:** OWL/RDF, Protégé, SPARQL reasoners |
34 | 44 |
|
35 | | -What exists (we don't build): |
36 | | -- **Syntactic:** Schema validators, Great Expectations, dbt tests, SQL constraints |
37 | | -- **Semantic:** OWL/RDF, Protégé, SPARQL reasoners |
| 45 | +## What We Build |
38 | 46 |
|
39 | | -What we build (pragmatics layer): |
40 | | -- Context items with latitude |
41 | | -- Thread traversal for query-relevant context |
42 | | -- Pack compilation for domain bundles |
43 | | -- Docstring injection for LLM grounding |
| 47 | +**Pragmatic:** Context items, latitude, thread traversal, pack compilation, docstring injection |
0 commit comments