File size: 4,126 Bytes
6256b19
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
"""Validator that keeps synthetic IES Turtle honest.
Rejects anything that (a) doesn't parse, (b) uses an ies: term not in the real
ontology, or (c) is too trivial to be a useful example."""
import json, pathlib, re
from rdflib import Graph, RDF, RDFS, Namespace

ROOT = pathlib.Path("/Users/fabio/projects/qwen-ies-ft")
IES = "http://ies.data.gov.uk/ontology/ies4#"
_vocab = json.loads((ROOT/"data"/"vocab.json").read_text())
REAL = set(_vocab["all_terms"])
CLASSES = set(_vocab["class_names"])

PREFIX_LINE = f"@prefix ies: <{IES}>"

def validate_turtle(ttl: str, min_triples: int = 4):
    if PREFIX_LINE not in ttl.replace("  ", " ").replace(" .", ""):
        # be lenient about spacing, but the ies namespace must be declared correctly
        if IES not in ttl:
            return False, "ies namespace not declared", 0, []
    g = Graph()
    try:
        g.parse(data=ttl, format="turtle")
    except Exception as e:
        return False, f"parse error: {str(e)[:120]}", 0, []
    n = len(g)
    if n < min_triples:
        return False, f"too few triples ({n})", n, []
    # every ies: term used must exist in the real ontology
    used = set()
    for t in g:
        for node in t:
            s = str(node)
            if s.startswith(IES):
                used.add(s.split("#")[-1])
    bad = sorted(used - REAL)
    if bad:
        return False, f"unknown ies terms: {bad[:6]}", n, sorted(used)
    # must instantiate at least one real IES class
    typed_classes = {str(o).split("#")[-1] for _, p, o in g.triples((None, RDF.type, None))
                     if str(o).startswith(IES)}
    if not (typed_classes & CLASSES):
        return False, "no rdf:type to a real IES class", n, sorted(used)
    return True, "ok", n, sorted(used)


# ---- structural conformance: domain/range checking with subclass closure ----
_PROPS = _vocab["properties"]
_SUPERS = {c: set(d["subClassOf"]) for c, d in _vocab["classes"].items()}

def _ancestors(c, seen=None):
    seen = seen or set()
    for s in _SUPERS.get(c, ()):
        if s not in seen:
            seen.add(s); _ancestors(s, seen)
    return seen

def _conforms(types, allowed):
    """any of the node's types is the allowed class or a descendant of it"""
    for t in types:
        if t in allowed: return True
        if _ancestors(t) & set(allowed): return True
    return False

def structural_conformance(ttl: str):
    """Fraction of checkable ies: property usages whose subject/object types satisfy the
    property's declared rdfs:domain/range (with subclass closure). Returns (ratio,
    checked, violations)."""
    g = Graph()
    try: g.parse(data=ttl, format="turtle")
    except Exception: return 0.0, 0, ["unparseable"]
    types = {}
    for s, _, o in g.triples((None, RDF.type, None)):
        if str(o).startswith(IES):
            types.setdefault(s, set()).add(str(o).split("#")[-1])
    checked = 0; bad = []
    for s, p, o in g:
        ps = str(p)
        if not ps.startswith(IES): continue
        pn = ps.split("#")[-1]
        d = _PROPS.get(pn)
        if not d: continue
        if d["domain"] and s in types:
            checked += 1
            if not _conforms(types[s], d["domain"]):
                bad.append(f"{pn}: subject {sorted(types[s])} not in domain {d['domain']}")
        if d["range"] and o in types:
            checked += 1
            if not _conforms(types[o], d["range"]):
                bad.append(f"{pn}: object {sorted(types[o])} not in range {d['range']}")
    ratio = 1.0 if checked == 0 else 1.0 - len(bad)/checked
    return ratio, checked, bad[:8]


if __name__ == "__main__":
    import sys
    # smoke test: a real sample must pass, a hallucinated term must fail
    good = (ROOT/"IES4"/"Sample Data"/"hospital.ttl").read_text()
    print("REAL hospital.ttl  ->", validate_turtle(good)[:3])
    bad = good + "\ndata:Fred ies:hasTelepathicLinkTo data:Ghost .\n"
    print("HALLUCINATED term  ->", validate_turtle(bad)[:2])
    broken = "@prefix ies: <http://ies.data.gov.uk/ontology/ies4#> .\ndata:x a ies:Person"  # no closing dot
    print("BROKEN syntax      ->", validate_turtle(broken)[:2])