Benjamin Admin
afb3f83f30
Engine precision (stop foreign-machine patterns leaking into a project):
- Wire project.MachineType into the engine machine-type gate (empty input no
longer fires every machine class — press/cnc/excavator/crane/medical...).
- Capability-domain gating extended by 7 domains (outdoor, ventilation,
machining, bulk, palletizer, playground, fitness) so domain-specific hazards
only fire when the narrative names that domain; emitted via keyword_dictionary.
- Relevance backstop moved into iace (single gating contract, testable), and its
dominant false-anchor class removed (a long pattern word no longer matches a
short common token; prepositions/leitung added to the generic stoplist).
- New guard tests: TestCrossDomainPrecision (full pipeline, 0 foreign per GT) and
TestPatternReachability now asserts 0 dead patterns. Both GTs keep coverage 1.0.
Reachability fix: the 51 dead patterns required electrical/pneumatic/hydraulic
tags nothing produced — renamed to the canonical electrical_energy/
pneumatic_pressure/hydraulic_pressure/hydraulic_part.
Component review (negation is best-effort + expert-correctable):
- Parser surfaces negated components (ComponentMatch.Negated) instead of dropping
them; negated contribute no tags/energy → no phantom hazards.
- presence_status (vorhanden|nicht_vorhanden|geloescht) + ce_marked on components;
only `vorhanden` feed matching. CE+safety-relevant flags the PL/SIL obligation.
- Force re-seed preserves the expert's component decisions instead of wiping them.
- Tag-based component→hazard assignment (was: all on the first component).
- Negation-aware narrative parsing ("keine Pneumatik" no longer extracts it).
Local-dev DB: ai-sdk sets search_path=compliance,core,public; reconcile migrations
152-156 bring the consolidated local iace tables to the current schema + add the
presence_status/ce_marked columns. Machine-type vocabulary endpoint for the form.
[migration-approved]
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
79 lines
2.8 KiB
Go
79 lines
2.8 KiB
Go
package iace
|
|
|
|
import "strings"
|
|
|
|
// Negation-aware keyword matching for the narrative parser.
|
|
//
|
|
// The limits form often states what a machine does NOT have, e.g.
|
|
// "Keine pneumatischen oder hydraulischen Schnittstellen. Hubantrieb ueber
|
|
// Kette." Naive substring matching wrongly extracted hydraulic/pneumatic
|
|
// components (and their energy sources), which then generated phantom hazards
|
|
// the assessor never raised. We only suppress DETERMINER negations
|
|
// ("keine/kein/ohne/weder <X>") — these directly negate the following noun
|
|
// phrase and are safe. Plain "nicht" is intentionally excluded: it modifies
|
|
// verbs/adjectives and over-negates ("Schutz nicht erforderlich, Zylinder
|
|
// vorhanden").
|
|
|
|
var negationDeterminers = map[string]bool{
|
|
"keine": true, "kein": true, "keinen": true, "keiner": true,
|
|
"keinem": true, "keines": true, "keinerlei": true, "weder": true,
|
|
"ohne": true, "no": true, "without": true,
|
|
}
|
|
|
|
// Tokens that end a negation's scope: a contrast ("aber") or a positive-presence
|
|
// cue ("mit ... vorhanden"). After one of these, a following keyword is positive.
|
|
var negationScopeEnders = map[string]bool{
|
|
"aber": true, "jedoch": true, "sondern": true, "doch": true,
|
|
"mit": true, "vorhanden": true, "verbaut": true, "vorgesehen": true,
|
|
"installiert": true, "ausgestattet": true, "but": true, "with": true,
|
|
}
|
|
|
|
// A negation determiner only reaches a keyword a few tokens away (a short list
|
|
// like "keine A, B oder C"). Beyond this span we assume the keyword is unrelated.
|
|
const negationMaxTokenSpan = 8
|
|
|
|
// keywordIsNegated reports whether the keyword occurrence starting at byte index
|
|
// idx in the (already normalised) text sits inside the scope of a determiner
|
|
// negation. It walks back to the start of the current sentence, then scans the
|
|
// preceding tokens right-to-left for a determiner negation, stopping at any
|
|
// scope-ender or after negationMaxTokenSpan tokens.
|
|
func keywordIsNegated(text string, idx int) bool {
|
|
start := 0
|
|
for i := idx - 1; i >= 0; i-- {
|
|
c := text[i]
|
|
if c == '.' || c == '\n' || c == ';' || c == '!' || c == '?' || c == ':' {
|
|
start = i + 1
|
|
break
|
|
}
|
|
}
|
|
tokens := strings.Fields(text[start:idx])
|
|
for d := 0; d < len(tokens) && d < negationMaxTokenSpan; d++ {
|
|
w := strings.Trim(tokens[len(tokens)-1-d], ",.;:()-")
|
|
if negationScopeEnders[w] {
|
|
return false
|
|
}
|
|
if negationDeterminers[w] {
|
|
return true
|
|
}
|
|
}
|
|
return false
|
|
}
|
|
|
|
// hasUnnegatedOccurrence reports whether kw appears in text at least once outside
|
|
// a negation scope. A term that is ONLY ever negated must not create components,
|
|
// energy sources or tags.
|
|
func hasUnnegatedOccurrence(text, kw string) bool {
|
|
from := 0
|
|
for {
|
|
rel := strings.Index(text[from:], kw)
|
|
if rel < 0 {
|
|
return false
|
|
}
|
|
abs := from + rel
|
|
if !keywordIsNegated(text, abs) {
|
|
return true
|
|
}
|
|
from = abs + len(kw)
|
|
}
|
|
}
|