File size: 10,904 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 | """Correct-by-construction IES4 training data.
Builds valid IES4 graphs with telicent-ies-tool across 12 patterns, double-validates
each (telicent SHACL + our term-membership validator), and emits
{pattern, facts, turtle} records. A deterministic natural-language 'facts' string is
attached; a later step paraphrases it with the local Qwen for linguistic diversity.
Run under .venv311 (telicent needs py>=3.10)."""
import sys, json, pathlib, random, argparse
sys.path.insert(0, str(pathlib.Path(__file__).parent))
from ies_tool.ies_tool import (IESTool, Person, Organisation, Event, Location,
Device, Asset, Communication, PartyInCommunication)
from iesval import validate_turtle
random.seed(101)
ROOT = pathlib.Path("/Users/fabio/projects/qwen-ies-ft")
GIVEN = ["Fred","Alice","Barry","Priya","Vladimir","Chen","Amara","Sofia","Liam","Yusuf",
"Grace","Diego","Nadia","Tom","Ivy","Omar","Hannah","Kofi","Elena","Sam"]
SUR = ["Smith","Okafor","Patel","Ivanov","Wang","Rossi","Khan","Nguyen","Brown","Silva",
"Adebayo","Muller","Kowalski","Andersson","Costa","Haddad","Murphy","Reyes"]
ORGS = ["Acme Logistics Ltd","Northgate Shipping","Meridian Bank","Halcyon Devices",
"Blue Harbour Freight","Sterling Manufacturing","Corvus Security","Pine Valley Clinic",
"Redwood Analytics","Orion Telecom","Harbourline Ferries","Vertex Engineering"]
PLACES = ["Port of Dover","Heathrow Terminal 4","Manchester Depot","University Hospital",
"Felixstowe Docks","Kings Cross Station","Birmingham Warehouse","Gatwick Cargo",
"Southampton Container Port","Leeds Distribution Centre","Bristol Harbour"]
CITIES = ["London","Manchester","Birmingham","Glasgow","Cardiff","Belfast","Leeds","Dover"]
DEVICES = ["Nokia handset","Samsung Galaxy phone","Toyota Hilux van","DAF articulated lorry",
"Dell laptop","Yamaha outboard motor","forklift truck","Land Rover Defender"]
POSTS = ["Warehouse Supervisor","Chief Engineer","Compliance Officer","Ship's Master",
"Security Analyst","Ward Nurse","Logistics Coordinator","Border Officer"]
UNITS = [("kilogram","kg"),("metre","m"),("year","yr"),("centimetre","cm")]
def yr(a=1970,b=2005): return random.randint(a,b)
def date(y=None):
y = y or yr(2010,2024)
return f"{y}-{random.randint(1,12):02d}-{random.randint(1,28):02d}"
def dt(): return date(yr(2018,2024))+f"T{random.randint(6,20):02d}:00:00"
def new():
t = IESTool(); t.default_data_namespace = "http://data.gov.uk/testdata#"; return t
# ---- each template returns (facts:str, turtle:str) using a fresh tool ----
def t_employment():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); org=random.choice(ORGS)
o=Organisation(tool=t, name=org)
dob=date(yr(1960,1998))
p=Person(tool=t, given_name=g, surname=s, date_of_birth=dob)
start=date(yr(2012,2020)); p.works_for(o, start=start)
facts=f"{g} {s}, born {dob}, has worked for {org} since {start}."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_employment_ended():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); org=random.choice(ORGS)
o=Organisation(tool=t, name=org)
p=Person(tool=t, given_name=g, surname=s)
a=yr(2008,2016); b=yr(2017,2023)
p.works_for(o, start=date(a), end=date(b))
facts=f"{g} {s} worked for {org} from {a} until {b}, then left."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_birth_death():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR)
bp=Location(tool=t); bp.add_name(random.choice(CITIES))
p=Person(tool=t, given_name=g, surname=s)
p.add_birth(date_of_birth=date(yr(1930,1960)), place_of_birth=bp)
if random.random()<0.6:
dp=Location(tool=t); dp.add_name(random.choice(CITIES))
p.add_death(date_of_death=date(yr(2000,2023)), place_of_death=dp)
facts=f"{g} {s} was born in one city and later died in another."
else:
facts=f"{g} {s} was born in a UK city."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_event_participation():
t=new(); loc=Location(tool=t); loc.add_name(random.choice(PLACES))
ev=Event(tool=t, start=dt(), end=dt()); ev.add_name(random.choice(
["team meeting","container inspection","cargo handover","security briefing","site visit"]))
ev.in_location(loc)
people=[]
for _ in range(random.randint(2,4)):
p=Person(tool=t, given_name=random.choice(GIVEN), surname=random.choice(SUR))
ev.add_participant(p); people.append(p)
facts=(f"{len(people)} people took part in an event at {loc}, which had a start and end time.")
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_identifier():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR)
p=Person(tool=t, given_name=g, surname=s)
num=f"{random.randint(10,99)}{random.choice('ABCDEFGH')}{random.randint(1000,9999)}"
p.add_identifier(num, id_class="http://ies.data.gov.uk/ontology/ies4#NationalIdentityNumber")
facts=f"{g} {s} is identified by the national identity number {num}."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_ownership():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); d=random.choice(DEVICES)
p=Person(tool=t, given_name=g, surname=s)
dev=Device(tool=t); dev.add_name(d)
p.owns(dev, start=date(yr(2015,2022)))
facts=f"{g} {s} owns a {d}, acquired at a known date."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_post_holding():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); org=random.choice(ORGS); post=random.choice(POSTS)
o=Organisation(tool=t, name=org)
pp=o.create_post(name=post, start=date(yr(2016,2021)))
p=Person(tool=t, given_name=g, surname=s)
p.in_post(pp, start=date(yr(2016,2021)))
facts=f"{g} {s} holds the post of {post} at {org}."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_location_state():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR)
loc=Location(tool=t); loc.add_name(random.choice(PLACES))
p=Person(tool=t, given_name=g, surname=s)
p.create_state(start=dt(), end=dt(), in_location=loc)
facts=f"{g} {s} was present at {loc} for a bounded period of time."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_access():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); d=random.choice(DEVICES)
p=Person(tool=t, given_name=g, surname=s)
dev=Device(tool=t); dev.add_name(d)
p.has_access_to(dev, start=date(yr(2019,2023)))
facts=f"{g} {s} has access to a {d} from a given date."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_org_owns_asset():
t=new(); org=random.choice(ORGS); d=random.choice(DEVICES)
o=Organisation(tool=t, name=org); a=Asset(tool=t); a.add_name(d)
o.owns(a, start=date(yr(2010,2020)))
facts=f"{org} owns a {d} as one of its assets."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_possession():
t=new(); g=random.choice(GIVEN); s=random.choice(SUR); d=random.choice(DEVICES)
p=Person(tool=t, given_name=g, surname=s)
dev=Device(tool=t); dev.add_name(d)
p.in_possession_of(dev, start=dt())
facts=f"{g} {s} was in possession of a {d} from a certain time."
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_communication():
t=new(); g1,s1=random.choice(GIVEN),random.choice(SUR); g2,s2=random.choice(GIVEN),random.choice(SUR)
p1=Person(tool=t, given_name=g1, surname=s1); p2=Person(tool=t, given_name=g2, surname=s2)
d1=Device(tool=t); d1.add_name(random.choice(DEVICES)); d2=Device(tool=t); d2.add_name(random.choice(DEVICES))
st=dt()
comm=Communication(tool=t, start=st)
pa=comm.create_party(); pa.add_person(p1); pa.add_device(d1)
pb=comm.create_party(); pb.add_person(p2); pb.add_device(d2)
facts=(f"{g1} {s1} and {g2} {s2} took part in a communication that began at {st}, "
f"each using their own device.")
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_composite_case():
# person + employment + attends event at location + possesses device
t=new(); g,s=random.choice(GIVEN),random.choice(SUR); org=random.choice(ORGS)
o=Organisation(tool=t, name=org)
p=Person(tool=t, given_name=g, surname=s, date_of_birth=date(yr(1965,1995)))
p.works_for(o, start=date(yr(2014,2020)))
dev=Device(tool=t); dev.add_name(random.choice(DEVICES)); p.in_possession_of(dev, start=dt())
loc=Location(tool=t); loc.add_name(random.choice(PLACES))
ev=Event(tool=t, start=dt(), end=dt()); ev.add_name(random.choice(
["site inspection","handover meeting","security review"])); ev.in_location(loc)
ev.add_participant(p)
facts=(f"{g} {s} works for {org}, was in possession of a device, and attended an event "
f"with a start and end time held at {loc}.")
return facts, t.get_rdf(rdf_format="turtle")["triples"]
def t_composite_org():
# organisation with post, post-holder, owned asset, and an event at a location
t=new(); org=random.choice(ORGS); post=random.choice(POSTS); g,s=random.choice(GIVEN),random.choice(SUR)
o=Organisation(tool=t, name=org)
pp=o.create_post(name=post, start=date(yr(2015,2020)))
p=Person(tool=t, given_name=g, surname=s); p.in_post(pp, start=date(yr(2016,2021)))
a=Asset(tool=t); a.add_name(random.choice(DEVICES)); o.owns(a, start=date(yr(2012,2019)))
loc=Location(tool=t); loc.add_name(random.choice(PLACES))
ev=Event(tool=t, start=dt(), end=dt()); ev.add_name("audit visit"); ev.in_location(loc)
ev.add_participant(p)
facts=(f"{org} has the post of {post}, held by {g} {s}, owns an asset, and hosted an "
f"audit visit at {loc} that {g} attended.")
return facts, t.get_rdf(rdf_format="turtle")["triples"]
TEMPLATES = [t_employment, t_employment_ended, t_birth_death, t_event_participation,
t_identifier, t_ownership, t_post_holding, t_location_state,
t_access, t_org_owns_asset, t_possession,
t_communication, t_composite_case, t_composite_org]
def main():
ap=argparse.ArgumentParser(); ap.add_argument("--n", type=int, default=1200)
args=ap.parse_args()
out=(ROOT/"data"/"ground.jsonl").open("w")
kept=0; fail=0; per={}
for i in range(args.n):
fn=TEMPLATES[i % len(TEMPLATES)]
try:
facts, ttl = fn()
except Exception as e:
fail+=1; continue
ok, reason, n, _ = validate_turtle(ttl, min_triples=5)
if ok:
out.write(json.dumps({"pattern":fn.__name__, "facts":facts, "turtle":ttl})+"\n")
kept+=1; per[fn.__name__]=per.get(fn.__name__,0)+1
else:
fail+=1
out.close()
print(f"ground pairs kept={kept} failed={fail}")
for k in sorted(per): print(f" {k}: {per[k]}")
if __name__=="__main__":
main()
|