spaCy/spacy/matcher.pyx
Matthew Honnibal 6a8cb905aa
Merge pull request #1876 from GregDubbin/master
Pattern matcher fixes
2018-01-24 16:38:11 +01:00

752 lines
28 KiB
Cython

# cython: profile=True
# cython: infer_types=True
# coding: utf8
from __future__ import unicode_literals
import ujson
from cymem.cymem cimport Pool
from preshed.maps cimport PreshMap
from libcpp.vector cimport vector
from libcpp.pair cimport pair
from cython.operator cimport dereference as deref
from murmurhash.mrmr cimport hash64
from libc.stdint cimport int32_t
# try:
# from libcpp.unordered_map cimport unordered_map as umap
# except:
# from libcpp.map cimport map as umap
from .typedefs cimport attr_t
from .typedefs cimport hash_t
from .structs cimport TokenC
from .tokens.doc cimport Doc, get_token_attr
from .vocab cimport Vocab
from .attrs import IDS
from .attrs cimport attr_id_t, ID, NULL_ATTR
from .attrs import FLAG61 as U_ENT
from .attrs import FLAG60 as B2_ENT
from .attrs import FLAG59 as B3_ENT
from .attrs import FLAG58 as B4_ENT
from .attrs import FLAG57 as B5_ENT
from .attrs import FLAG56 as B6_ENT
from .attrs import FLAG55 as B7_ENT
from .attrs import FLAG54 as B8_ENT
from .attrs import FLAG53 as B9_ENT
from .attrs import FLAG52 as B10_ENT
from .attrs import FLAG51 as I3_ENT
from .attrs import FLAG50 as I4_ENT
from .attrs import FLAG49 as I5_ENT
from .attrs import FLAG48 as I6_ENT
from .attrs import FLAG47 as I7_ENT
from .attrs import FLAG46 as I8_ENT
from .attrs import FLAG45 as I9_ENT
from .attrs import FLAG44 as I10_ENT
from .attrs import FLAG43 as L2_ENT
from .attrs import FLAG42 as L3_ENT
from .attrs import FLAG41 as L4_ENT
from .attrs import FLAG40 as L5_ENT
from .attrs import FLAG39 as L6_ENT
from .attrs import FLAG38 as L7_ENT
from .attrs import FLAG37 as L8_ENT
from .attrs import FLAG36 as L9_ENT
from .attrs import FLAG35 as L10_ENT
cpdef enum quantifier_t:
_META
ONE
ZERO
ZERO_ONE
ZERO_PLUS
cdef enum action_t:
REJECT
ADVANCE
REPEAT
ACCEPT
ADVANCE_ZERO
ADVANCE_PLUS
ACCEPT_PREV
PANIC
# Each token pattern consists of a quantifier and 0+ (attr, value) pairs.
# A state is an (int, pattern pointer) pair, where the int is the start
# position, and the pattern pointer shows where we're up to
# in the pattern.
cdef struct AttrValueC:
attr_id_t attr
attr_t value
cdef struct TokenPatternC:
AttrValueC* attrs
int32_t nr_attr
quantifier_t quantifier
ctypedef TokenPatternC* TokenPatternC_ptr
# ctypedef pair[int, TokenPatternC_ptr] StateC
# Match Dictionary entry type
cdef struct MatchEntryC:
int32_t start
int32_t end
int32_t offset
# A state instance represents the information that defines a
# partial match
# start: the index of the first token in the partial match
# pattern: a pointer to the current token pattern in the full
# pattern
# last_match: The entry of the last span matched by the
# same pattern
cdef struct StateC:
int32_t start
TokenPatternC_ptr pattern
MatchEntryC* last_match
cdef TokenPatternC* init_pattern(Pool mem, attr_t entity_id,
object token_specs) except NULL:
pattern = <TokenPatternC*>mem.alloc(len(token_specs) + 1, sizeof(TokenPatternC))
cdef int i
for i, (quantifier, spec) in enumerate(token_specs):
pattern[i].quantifier = quantifier
pattern[i].attrs = <AttrValueC*>mem.alloc(len(spec), sizeof(AttrValueC))
pattern[i].nr_attr = len(spec)
for j, (attr, value) in enumerate(spec):
pattern[i].attrs[j].attr = attr
pattern[i].attrs[j].value = value
i = len(token_specs)
pattern[i].attrs = <AttrValueC*>mem.alloc(2, sizeof(AttrValueC))
pattern[i].attrs[0].attr = ID
pattern[i].attrs[0].value = entity_id
pattern[i].nr_attr = 0
return pattern
cdef attr_t get_pattern_key(const TokenPatternC* pattern) except 0:
while pattern.nr_attr != 0:
pattern += 1
id_attr = pattern[0].attrs[0]
assert id_attr.attr == ID
return id_attr.value
cdef int get_action(const TokenPatternC* pattern, const TokenC* token) nogil:
lookahead = &pattern[1]
for attr in pattern.attrs[:pattern.nr_attr]:
if get_token_attr(token, attr.attr) != attr.value:
if pattern.quantifier == ONE:
return REJECT
elif pattern.quantifier == ZERO:
return ACCEPT if lookahead.nr_attr == 0 else ADVANCE
elif pattern.quantifier in (ZERO_ONE, ZERO_PLUS):
return ACCEPT_PREV if lookahead.nr_attr == 0 else ADVANCE_ZERO
else:
return PANIC
if pattern.quantifier == ZERO:
return REJECT
elif lookahead.nr_attr == 0:
if pattern.quantifier == ZERO_PLUS:
return REPEAT
else:
return ACCEPT
elif pattern.quantifier in (ONE, ZERO_ONE):
return ADVANCE
elif pattern.quantifier == ZERO_PLUS:
# This is a bandaid over the 'shadowing' problem described here:
# https://github.com/explosion/spaCy/issues/864
next_action = get_action(lookahead, token)
if next_action is REJECT:
return REPEAT
else:
return ADVANCE_PLUS
else:
return PANIC
def _convert_strings(token_specs, string_store):
# Support 'syntactic sugar' operator '+', as combination of ONE, ZERO_PLUS
operators = {'!': (ZERO,), '*': (ZERO_PLUS,), '+': (ONE, ZERO_PLUS),
'?': (ZERO_ONE,), '1': (ONE,)}
tokens = []
op = ONE
for spec in token_specs:
if not spec:
# Signifier for 'any token'
tokens.append((ONE, [(NULL_ATTR, 0)]))
continue
token = []
ops = (ONE,)
for attr, value in spec.items():
if isinstance(attr, basestring) and attr.upper() == 'OP':
if value in operators:
ops = operators[value]
else:
msg = "Unknown operator '%s'. Options: %s"
raise KeyError(msg % (value, ', '.join(operators.keys())))
if isinstance(attr, basestring):
attr = IDS.get(attr.upper())
if isinstance(value, basestring):
value = string_store.add(value)
if isinstance(value, bool):
value = int(value)
if attr is not None:
token.append((attr, value))
for op in ops:
tokens.append((op, token))
return tokens
def merge_phrase(matcher, doc, i, matches):
"""Callback to merge a phrase on match."""
ent_id, label, start, end = matches[i]
span = doc[start:end]
span.merge(ent_type=label, ent_id=ent_id)
def unpickle_matcher(vocab, patterns, callbacks):
matcher = Matcher(vocab)
for key, specs in patterns.items():
callback = callbacks.get(key, None)
matcher.add(key, callback, *specs)
return matcher
cdef class Matcher:
"""Match sequences of tokens, based on pattern rules."""
cdef Pool mem
cdef vector[TokenPatternC*] patterns
cdef readonly Vocab vocab
cdef public object _patterns
cdef public object _entities
cdef public object _callbacks
def __init__(self, vocab):
"""Create the Matcher.
vocab (Vocab): The vocabulary object, which must be shared with the
documents the matcher will operate on.
RETURNS (Matcher): The newly constructed object.
"""
self._patterns = {}
self._entities = {}
self._callbacks = {}
self.vocab = vocab
self.mem = Pool()
def __reduce__(self):
data = (self.vocab, self._patterns, self._callbacks)
return (unpickle_matcher, data, None, None)
def __len__(self):
"""Get the number of rules added to the matcher. Note that this only
returns the number of rules (identical with the number of IDs), not the
number of individual patterns.
RETURNS (int): The number of rules.
"""
return len(self._patterns)
def __contains__(self, key):
"""Check whether the matcher contains rules for a match ID.
key (unicode): The match ID.
RETURNS (bool): Whether the matcher contains rules for this match ID.
"""
return self._normalize_key(key) in self._patterns
def add(self, key, on_match, *patterns):
"""Add a match-rule to the matcher. A match-rule consists of: an ID
key, an on_match callback, and one or more patterns.
If the key exists, the patterns are appended to the previous ones, and
the previous on_match callback is replaced. The `on_match` callback
will receive the arguments `(matcher, doc, i, matches)`. You can also
set `on_match` to `None` to not perform any actions.
A pattern consists of one or more `token_specs`, where a `token_spec`
is a dictionary mapping attribute IDs to values, and optionally a
quantifier operator under the key "op". The available quantifiers are:
'!': Negate the pattern, by requiring it to match exactly 0 times.
'?': Make the pattern optional, by allowing it to match 0 or 1 times.
'+': Require the pattern to match 1 or more times.
'*': Allow the pattern to zero or more times.
The + and * operators are usually interpretted "greedily", i.e. longer
matches are returned where possible. However, if you specify two '+'
and '*' patterns in a row and their matches overlap, the first
operator will behave non-greedily. This quirk in the semantics makes
the matcher more efficient, by avoiding the need for back-tracking.
key (unicode): The match ID.
on_match (callable): Callback executed on match.
*patterns (list): List of token descritions.
"""
for pattern in patterns:
if len(pattern) == 0:
msg = ("Cannot add pattern for zero tokens to matcher.\n"
"key: {key}\n")
raise ValueError(msg.format(key=key))
key = self._normalize_key(key)
for pattern in patterns:
specs = _convert_strings(pattern, self.vocab.strings)
self.patterns.push_back(init_pattern(self.mem, key, specs))
self._patterns.setdefault(key, [])
self._callbacks[key] = on_match
self._patterns[key].extend(patterns)
def remove(self, key):
"""Remove a rule from the matcher. A KeyError is raised if the key does
not exist.
key (unicode): The ID of the match rule.
"""
key = self._normalize_key(key)
self._patterns.pop(key)
self._callbacks.pop(key)
cdef int i = 0
while i < self.patterns.size():
pattern_key = get_pattern_key(self.patterns.at(i))
if pattern_key == key:
self.patterns.erase(self.patterns.begin()+i)
else:
i += 1
def has_key(self, key):
"""Check whether the matcher has a rule with a given key.
key (string or int): The key to check.
RETURNS (bool): Whether the matcher has the rule.
"""
key = self._normalize_key(key)
return key in self._patterns
def get(self, key, default=None):
"""Retrieve the pattern stored for a key.
key (unicode or int): The key to retrieve.
RETURNS (tuple): The rule, as an (on_match, patterns) tuple.
"""
key = self._normalize_key(key)
if key not in self._patterns:
return default
return (self._callbacks[key], self._patterns[key])
def pipe(self, docs, batch_size=1000, n_threads=2):
"""Match a stream of documents, yielding them in turn.
docs (iterable): A stream of documents.
batch_size (int): Number of documents to accumulate into a working set.
n_threads (int): The number of threads with which to work on the buffer
in parallel, if the implementation supports multi-threading.
YIELDS (Doc): Documents, in order.
"""
for doc in docs:
self(doc)
yield doc
def __call__(self, Doc doc):
"""Find all token sequences matching the supplied pattern.
doc (Doc): The document to match over.
RETURNS (list): A list of `(key, start, end)` tuples,
describing the matches. A match tuple describes a span
`doc[start:end]`. The `label_id` and `key` are both integers.
"""
cdef vector[StateC] partials
cdef int n_partials = 0
cdef int q = 0
cdef int i, token_i
cdef const TokenC* token
cdef StateC state
cdef int j = 0
cdef int k
cdef bint overlap = False
cdef MatchEntryC* state_match
cdef MatchEntryC* last_matches = <MatchEntryC*>self.mem.alloc(self.patterns.size(),sizeof(MatchEntryC))
for i in range(self.patterns.size()):
last_matches[i].start = 0
last_matches[i].end = 0
last_matches[i].offset = 0
matches = []
for token_i in range(doc.length):
token = &doc.c[token_i]
q = 0
# Go over the open matches, extending or finalizing if able.
# Otherwise, we over-write them (q doesn't advance)
#for state in partials:
j=0
while j < n_partials:
state = partials[j]
action = get_action(state.pattern, token)
j += 1
# Skip patterns that would overlap with an existing match
# Patterns overlap an existing match if they point to the
# same final state and start between the start and end
# of said match.
# Different patterns with the same label are allowed to
# overlap.
state_match = state.last_match
if (state.start > state_match.start
and state.start < state_match.end):
continue
if action == PANIC:
raise Exception("Error selecting action in matcher")
while action == ADVANCE_ZERO:
state.pattern += 1
action = get_action(state.pattern, token)
if action == PANIC:
raise Exception("Error selecting action in matcher")
# ADVANCE_PLUS acts like REPEAT, but also pushes a partial that
# acts like and ADVANCE_ZERO
if action == ADVANCE_PLUS:
state.pattern += 1
partials.push_back(state)
n_partials += 1
state.pattern -= 1
action = REPEAT
if action == ADVANCE:
state.pattern += 1
# Check for partial matches that are at the same spec in the same pattern
# Keep the longer of the matches
# This ensures that there are never more then 2 partials for every spec
# in a pattern (one of which gets pruned in this step)
overlap=False
for i in range(q):
if state.pattern == partials[i].pattern and state.start < partials[i].start:
partials[i] = state
j = i
overlap = True
break
if overlap:
continue
overlap=False
for i in range(q):
if state.pattern == partials[i].pattern:
overlap = True
break
if overlap:
continue
if action == REPEAT:
# Leave the state in the queue, and advance to next slot
# (i.e. we don't overwrite -- we want to greedily match
# more pattern.
partials[q] = state
q += 1
elif action == REJECT:
pass
elif action == ADVANCE:
partials[q] = state
q += 1
elif action in (ACCEPT, ACCEPT_PREV):
# TODO: What to do about patterns starting with ZERO? Need
# to adjust the start position.
start = state.start
end = token_i+1 if action == ACCEPT else token_i
ent_id = state.pattern[1].attrs[0].value
label = state.pattern[1].attrs[1].value
# Check that this match doesn't overlap with an earlier match.
# Only overwrite an earlier match if it is a substring of this
# match (i.e. it starts after this match starts).
state_match = state.last_match
if start >= state_match.end:
state_match.start = start
state_match.end = end
state_match.offset = len(matches)
matches.append((ent_id,start,end))
elif start <= state_match.start and end >= state_match.end:
if len(matches) == 0:
assert state_match.offset==0
state_match.offset = 0
matches.append((ent_id,start,end))
else:
i = state_match.offset
matches[i] = (ent_id,start,end)
state_match.start = start
state_match.end = end
else:
pass
partials.resize(q)
n_partials = q
# Check whether we open any new patterns on this token
i=0
for pattern in self.patterns:
# Skip patterns that would overlap with an existing match
# state_match = pattern.last_match
state_match = &last_matches[i]
i+=1
if (token_i > state_match.start
and token_i < state_match.end):
continue
action = get_action(pattern, token)
if action == PANIC:
raise Exception("Error selecting action in matcher")
while action in (ADVANCE_PLUS,ADVANCE_ZERO):
if action == ADVANCE_PLUS:
state.start = token_i
state.pattern = pattern
state.last_match = state_match
partials.push_back(state)
n_partials += 1
pattern += 1
action = get_action(pattern, token)
if action == ADVANCE:
pattern += 1
j=0
overlap = False
for j in range(q):
if pattern == partials[j].pattern:
overlap = True
break
if overlap:
continue
if action == REPEAT:
state.start = token_i
state.pattern = pattern
state.last_match = state_match
partials.push_back(state)
n_partials += 1
elif action == ADVANCE:
# TODO: What to do about patterns starting with ZERO? Need
# to adjust the start position.
state.start = token_i
state.pattern = pattern
state.last_match = state_match
partials.push_back(state)
n_partials += 1
elif action in (ACCEPT, ACCEPT_PREV):
start = token_i
end = token_i+1 if action == ACCEPT else token_i
ent_id = pattern[1].attrs[0].value
label = pattern[1].attrs[1].value
if start >= state_match.end:
state_match.start = start
state_match.end = end
state_match.offset = len(matches)
matches.append((ent_id,start,end))
if start <= state_match.start and end >= state_match.end:
if len(matches) == 0:
state_match.offset = 0
matches.append((ent_id,start,end))
else:
j = state_match.offset
matches[j] = (ent_id,start,end)
state_match.start = start
state_match.end = end
else:
pass
# Look for open patterns that are actually satisfied
for state in partials:
while state.pattern.quantifier in (ZERO, ZERO_ONE, ZERO_PLUS):
state.pattern += 1
if state.pattern.nr_attr == 0:
start = state.start
end = len(doc)
ent_id = state.pattern.attrs[0].value
label = state.pattern.attrs[1].value
state_match = state.last_match
if start >= state_match.end:
state_match.start = start
state_match.end = end
state_match.offset = len(matches)
matches.append((ent_id,start,end))
if start <= state_match.start and end >= state_match.end:
j = state_match.offset
if len(matches) == 0:
state_match.offset = 0
matches.append((ent_id,start,end))
else:
matches[j] = (ent_id,start,end)
state_match.start = start
state_match.end = end
else:
pass
for i, (ent_id, start, end) in enumerate(matches):
on_match = self._callbacks.get(ent_id)
if on_match is not None:
on_match(self, doc, i, matches)
return matches
def _normalize_key(self, key):
if isinstance(key, basestring):
return self.vocab.strings.add(key)
else:
return key
def get_bilou(length):
if length == 1:
return [U_ENT]
elif length == 2:
return [B2_ENT, L2_ENT]
elif length == 3:
return [B3_ENT, I3_ENT, L3_ENT]
elif length == 4:
return [B4_ENT, I4_ENT, I4_ENT, L4_ENT]
elif length == 5:
return [B5_ENT, I5_ENT, I5_ENT, I5_ENT, L5_ENT]
elif length == 6:
return [B6_ENT, I6_ENT, I6_ENT, I6_ENT, I6_ENT, L6_ENT]
elif length == 7:
return [B7_ENT, I7_ENT, I7_ENT, I7_ENT, I7_ENT, I7_ENT, L7_ENT]
elif length == 8:
return [B8_ENT, I8_ENT, I8_ENT, I8_ENT, I8_ENT, I8_ENT, I8_ENT, L8_ENT]
elif length == 9:
return [B9_ENT, I9_ENT, I9_ENT, I9_ENT, I9_ENT, I9_ENT, I9_ENT, I9_ENT,
L9_ENT]
elif length == 10:
return [B10_ENT, I10_ENT, I10_ENT, I10_ENT, I10_ENT, I10_ENT, I10_ENT,
I10_ENT, I10_ENT, L10_ENT]
else:
raise ValueError("Max length currently 10 for phrase matching")
cdef class PhraseMatcher:
cdef Pool mem
cdef Vocab vocab
cdef Matcher matcher
cdef PreshMap phrase_ids
cdef int max_length
cdef attr_t* _phrase_key
cdef public object _callbacks
cdef public object _patterns
def __init__(self, Vocab vocab, max_length=10):
self.mem = Pool()
self._phrase_key = <attr_t*>self.mem.alloc(max_length, sizeof(attr_t))
self.max_length = max_length
self.vocab = vocab
self.matcher = Matcher(self.vocab)
self.phrase_ids = PreshMap()
abstract_patterns = []
for length in range(1, max_length):
abstract_patterns.append([{tag: True}
for tag in get_bilou(length)])
self.matcher.add('Candidate', None, *abstract_patterns)
self._callbacks = {}
def __len__(self):
"""Get the number of rules added to the matcher. Note that this only
returns the number of rules (identical with the number of IDs), not the
number of individual patterns.
RETURNS (int): The number of rules.
"""
return len(self.phrase_ids)
def __contains__(self, key):
"""Check whether the matcher contains rules for a match ID.
key (unicode): The match ID.
RETURNS (bool): Whether the matcher contains rules for this match ID.
"""
cdef hash_t ent_id = self.matcher._normalize_key(key)
return ent_id in self._callbacks
def __reduce__(self):
return (self.__class__, (self.vocab,), None, None)
def add(self, key, on_match, *docs):
"""Add a match-rule to the matcher. A match-rule consists of: an ID
key, an on_match callback, and one or more patterns.
key (unicode): The match ID.
on_match (callable): Callback executed on match.
*docs (Doc): `Doc` objects representing match patterns.
"""
cdef Doc doc
for doc in docs:
if len(doc) >= self.max_length:
msg = (
"Pattern length (%d) >= phrase_matcher.max_length (%d). "
"Length can be set on initialization, up to 10."
)
raise ValueError(msg % (len(doc), self.max_length))
cdef hash_t ent_id = self.matcher._normalize_key(key)
self._callbacks[ent_id] = on_match
cdef int length
cdef int i
cdef hash_t phrase_hash
for doc in docs:
length = doc.length
tags = get_bilou(length)
for i in range(self.max_length):
self._phrase_key[i] = 0
for i, tag in enumerate(tags):
lexeme = self.vocab[doc.c[i].lex.orth]
lexeme.set_flag(tag, True)
self._phrase_key[i] = lexeme.orth
phrase_hash = hash64(self._phrase_key,
self.max_length * sizeof(attr_t), 0)
self.phrase_ids.set(phrase_hash, <void*>ent_id)
def __call__(self, Doc doc):
"""Find all sequences matching the supplied patterns on the `Doc`.
doc (Doc): The document to match over.
RETURNS (list): A list of `(key, start, end)` tuples,
describing the matches. A match tuple describes a span
`doc[start:end]`. The `label_id` and `key` are both integers.
"""
matches = []
for _, start, end in self.matcher(doc):
ent_id = self.accept_match(doc, start, end)
if ent_id is not None:
matches.append((ent_id, start, end))
for i, (ent_id, start, end) in enumerate(matches):
on_match = self._callbacks.get(ent_id)
if on_match is not None:
on_match(self, doc, i, matches)
return matches
def pipe(self, stream, batch_size=1000, n_threads=2):
"""Match a stream of documents, yielding them in turn.
docs (iterable): A stream of documents.
batch_size (int): Number of documents to accumulate into a working set.
n_threads (int): The number of threads with which to work on the buffer
in parallel, if the implementation supports multi-threading.
YIELDS (Doc): Documents, in order.
"""
for doc in stream:
self(doc)
yield doc
def accept_match(self, Doc doc, int start, int end):
assert (end - start) < self.max_length
cdef int i, j
for i in range(self.max_length):
self._phrase_key[i] = 0
for i, j in enumerate(range(start, end)):
self._phrase_key[i] = doc.c[j].lex.orth
cdef hash_t key = hash64(self._phrase_key,
self.max_length * sizeof(attr_t), 0)
ent_id = <hash_t>self.phrase_ids.get(key)
if ent_id == 0:
return None
else:
return ent_id