spaCy/spacy/kb.pyx

# cython: infer_types=True, profile=True
from typing import Iterator, Iterable, Callable, Dict, Any

import srsly
from cymem.cymem cimport Pool
from preshed.maps cimport PreshMap
from cpython.exc cimport PyErr_SetFromErrno
from libc.stdio cimport fopen, fclose, fread, fwrite, feof, fseek
from libc.stdint cimport int32_t, int64_t
from libcpp.vector cimport vector

from pathlib import Path
import warnings

from .typedefs cimport hash_t
from .errors import Errors, Warnings
from . import util
from .util import SimpleFrozenList, ensure_path

cdef class Candidate:
    """A `Candidate` object refers to a textual mention (`alias`) that may or may not be resolved
    to a specific `entity` from a Knowledge Base. This will be used as input for the entity linking
    algorithm which will disambiguate the various candidates to the correct one.
    Each candidate (alias, entity) pair is assigned to a certain prior probability.

    DOCS: https://spacy.io/api/kb/#candidate_init
    """

    def __init__(self, KnowledgeBase kb, entity_hash, entity_freq, entity_vector, alias_hash, prior_prob):
        self.kb = kb
        self.entity_hash = entity_hash
        self.entity_freq = entity_freq
        self.entity_vector = entity_vector
        self.alias_hash = alias_hash
        self.prior_prob = prior_prob

    @property
    def entity(self):
        """RETURNS (uint64): hash of the entity's KB ID/name"""
        return self.entity_hash

    @property
    def entity_(self):
        """RETURNS (str): ID/name of this entity in the KB"""
        return self.kb.vocab.strings[self.entity_hash]

    @property
    def alias(self):
        """RETURNS (uint64): hash of the alias"""
        return self.alias_hash

    @property
    def alias_(self):
        """RETURNS (str): ID of the original alias"""
        return self.kb.vocab.strings[self.alias_hash]

    @property
    def entity_freq(self):
        return self.entity_freq

    @property
    def entity_vector(self):
        return self.entity_vector

    @property
    def prior_prob(self):
        return self.prior_prob


def get_candidates(KnowledgeBase kb, span) -> Iterator[Candidate]:
    """
    Return candidate entities for a given span by using the text of the span as the alias
    and fetching appropriate entries from the index.
    This particular function is optimized to work with the built-in KB functionality,
    but any other custom candidate generation method can be used in combination with the KB as well.
    """
    return kb.get_alias_candidates(span.text)


cdef class KnowledgeBase:
    """A `KnowledgeBase` instance stores unique identifiers for entities and their textual aliases,
    to support entity linking of named entities to real-world concepts.

    DOCS: https://spacy.io/api/kb
    """

    def __init__(self, Vocab vocab, entity_vector_length):
        """Create a KnowledgeBase."""
        self.mem = Pool()
        self.entity_vector_length = entity_vector_length
        self._entry_index = PreshMap()
        self._alias_index = PreshMap()
        self.vocab = vocab
        self._create_empty_vectors(dummy_hash=self.vocab.strings[""])

    def _initialize_entities(self, int64_t nr_entities):
        self._entry_index = PreshMap(nr_entities + 1)
        self._entries = entry_vec(nr_entities + 1)

    def _initialize_vectors(self, int64_t nr_entities):
        self._vectors_table = float_matrix(nr_entities + 1)

    def _initialize_aliases(self, int64_t nr_aliases):
        self._alias_index = PreshMap(nr_aliases + 1)
        self._aliases_table = alias_vec(nr_aliases + 1)

    @property
    def entity_vector_length(self):
        """RETURNS (uint64): length of the entity vectors"""
        return self.entity_vector_length

    def __len__(self):
        return self.get_size_entities()

    def get_size_entities(self):
        return len(self._entry_index)

    def get_entity_strings(self):
        return [self.vocab.strings[x] for x in self._entry_index]

    def get_size_aliases(self):
        return len(self._alias_index)

    def get_alias_strings(self):
        return [self.vocab.strings[x] for x in self._alias_index]

    def add_entity(self, str entity, float freq, vector[float] entity_vector):
        """
        Add an entity to the KB, optionally specifying its log probability based on corpus frequency
        Return the hash of the entity ID/name at the end.
        """
        cdef hash_t entity_hash = self.vocab.strings.add(entity)

        # Return if this entity was added before
        if entity_hash in self._entry_index:
            warnings.warn(Warnings.W018.format(entity=entity))
            return

        # Raise an error if the provided entity vector is not of the correct length
        if len(entity_vector) != self.entity_vector_length:
            raise ValueError(Errors.E141.format(found=len(entity_vector), required=self.entity_vector_length))

        vector_index = self.c_add_vector(entity_vector=entity_vector)

        new_index = self.c_add_entity(entity_hash=entity_hash,
                                      freq=freq,
                                      vector_index=vector_index,
                                      feats_row=-1)  # Features table currently not implemented
        self._entry_index[entity_hash] = new_index

        return entity_hash

    cpdef set_entities(self, entity_list, freq_list, vector_list):
        if len(entity_list) != len(freq_list) or len(entity_list) != len(vector_list):
            raise ValueError(Errors.E140)

        nr_entities = len(set(entity_list))
        self._initialize_entities(nr_entities)
        self._initialize_vectors(nr_entities)

        i = 0
        cdef KBEntryC entry
        cdef hash_t entity_hash
        while i < len(entity_list):
            # only process this entity if its unique ID hadn't been added before
            entity_hash = self.vocab.strings.add(entity_list[i])
            if entity_hash in self._entry_index:
                warnings.warn(Warnings.W018.format(entity=entity_list[i]))

            else:
                entity_vector = vector_list[i]
                if len(entity_vector) != self.entity_vector_length:
                    raise ValueError(Errors.E141.format(found=len(entity_vector), required=self.entity_vector_length))

                entry.entity_hash = entity_hash
                entry.freq = freq_list[i]

                self._vectors_table[i] = entity_vector
                entry.vector_index = i

                entry.feats_row = -1   # Features table currently not implemented

                self._entries[i+1] = entry
                self._entry_index[entity_hash] = i+1

            i += 1

    def contains_entity(self, str entity):
        cdef hash_t entity_hash = self.vocab.strings.add(entity)
        return entity_hash in self._entry_index

    def contains_alias(self, str alias):
        cdef hash_t alias_hash = self.vocab.strings.add(alias)
        return alias_hash in self._alias_index

    def add_alias(self, str alias, entities, probabilities):
        """
        For a given alias, add its potential entities and prior probabilies to the KB.
        Return the alias_hash at the end
        """
        if alias is None or len(alias) == 0:
            raise ValueError(Errors.E890.format(alias=alias))

        previous_alias_nr = self.get_size_aliases()
        # Throw an error if the length of entities and probabilities are not the same
        if not len(entities) == len(probabilities):
            raise ValueError(Errors.E132.format(alias=alias,
                                                entities_length=len(entities),
                                                probabilities_length=len(probabilities)))

        # Throw an error if the probabilities sum up to more than 1 (allow for some rounding errors)
        prob_sum = sum(probabilities)
        if prob_sum > 1.00001:
            raise ValueError(Errors.E133.format(alias=alias, sum=prob_sum))

        cdef hash_t alias_hash = self.vocab.strings.add(alias)

        # Check whether this alias was added before
        if alias_hash in self._alias_index:
            warnings.warn(Warnings.W017.format(alias=alias))
            return

        cdef vector[int64_t] entry_indices
        cdef vector[float] probs

        for entity, prob in zip(entities, probabilities):
            entity_hash = self.vocab.strings[entity]
            if not entity_hash in self._entry_index:
                raise ValueError(Errors.E134.format(entity=entity))

            entry_index = <int64_t>self._entry_index.get(entity_hash)
            entry_indices.push_back(int(entry_index))
            probs.push_back(float(prob))

        new_index = self.c_add_aliases(alias_hash=alias_hash, entry_indices=entry_indices, probs=probs)
        self._alias_index[alias_hash] = new_index

        if previous_alias_nr + 1 != self.get_size_aliases():
            raise RuntimeError(Errors.E891.format(alias=alias))
        return alias_hash

    def append_alias(self, str alias, str entity, float prior_prob, ignore_warnings=False):
        """
        For an alias already existing in the KB, extend its potential entities with one more.
        Throw a warning if either the alias or the entity is unknown,
        or when the combination is already previously recorded.
        Throw an error if this entity+prior prob would exceed the sum of 1.
        For efficiency, it's best to use the method `add_alias` as much as possible instead of this one.
        """
        # Check if the alias exists in the KB
        cdef hash_t alias_hash = self.vocab.strings[alias]
        if not alias_hash in self._alias_index:
            raise ValueError(Errors.E176.format(alias=alias))

        # Check if the entity exists in the KB
        cdef hash_t entity_hash = self.vocab.strings[entity]
        if not entity_hash in self._entry_index:
            raise ValueError(Errors.E134.format(entity=entity))
        entry_index = <int64_t>self._entry_index.get(entity_hash)

        # Throw an error if the prior probabilities (including the new one) sum up to more than 1
        alias_index = <int64_t>self._alias_index.get(alias_hash)
        alias_entry = self._aliases_table[alias_index]
        current_sum = sum([p for p in alias_entry.probs])
        new_sum = current_sum + prior_prob

        if new_sum > 1.00001:
            raise ValueError(Errors.E133.format(alias=alias, sum=new_sum))

        entry_indices = alias_entry.entry_indices

        is_present = False
        for i in range(entry_indices.size()):
            if entry_indices[i] == int(entry_index):
                is_present = True

        if is_present:
            if not ignore_warnings:
                warnings.warn(Warnings.W024.format(entity=entity, alias=alias))
        else:
            entry_indices.push_back(int(entry_index))
            alias_entry.entry_indices = entry_indices

            probs = alias_entry.probs
            probs.push_back(float(prior_prob))
            alias_entry.probs = probs
            self._aliases_table[alias_index] = alias_entry

    def get_alias_candidates(self, str alias) -> Iterator[Candidate]:
        """
        Return candidate entities for an alias. Each candidate defines the entity, the original alias,
        and the prior probability of that alias resolving to that entity.
        If the alias is not known in the KB, and empty list is returned.
        """
        cdef hash_t alias_hash = self.vocab.strings[alias]
        if not alias_hash in self._alias_index:
            return []
        alias_index = <int64_t>self._alias_index.get(alias_hash)
        alias_entry = self._aliases_table[alias_index]

        return [Candidate(kb=self,
                          entity_hash=self._entries[entry_index].entity_hash,
                          entity_freq=self._entries[entry_index].freq,
                          entity_vector=self._vectors_table[self._entries[entry_index].vector_index],
                          alias_hash=alias_hash,
                          prior_prob=prior_prob)
                for (entry_index, prior_prob) in zip(alias_entry.entry_indices, alias_entry.probs)
                if entry_index != 0]

    def get_vector(self, str entity):
        cdef hash_t entity_hash = self.vocab.strings[entity]

        # Return an empty list if this entity is unknown in this KB
        if entity_hash not in self._entry_index:
            return [0] * self.entity_vector_length
        entry_index = self._entry_index[entity_hash]

        return self._vectors_table[self._entries[entry_index].vector_index]

    def get_prior_prob(self, str entity, str alias):
        """ Return the prior probability of a given alias being linked to a given entity,
        or return 0.0 when this combination is not known in the knowledge base"""
        cdef hash_t alias_hash = self.vocab.strings[alias]
        cdef hash_t entity_hash = self.vocab.strings[entity]

        if entity_hash not in self._entry_index or alias_hash not in self._alias_index:
            return 0.0

        alias_index = <int64_t>self._alias_index.get(alias_hash)
        entry_index = self._entry_index[entity_hash]

        alias_entry = self._aliases_table[alias_index]
        for (entry_index, prior_prob) in zip(alias_entry.entry_indices, alias_entry.probs):
            if self._entries[entry_index].entity_hash == entity_hash:
                return prior_prob

        return 0.0

    def to_bytes(self, **kwargs):
        """Serialize the current state to a binary string.
        """
        def serialize_header():
            header = (self.get_size_entities(), self.get_size_aliases(), self.entity_vector_length)
            return srsly.json_dumps(header)

        def serialize_entries():
            i = 1
            tuples = []
            for entry_hash, entry_index in sorted(self._entry_index.items(), key=lambda x: x[1]):
                entry = self._entries[entry_index]
                assert entry.entity_hash == entry_hash
                assert entry_index == i
                tuples.append((entry.entity_hash, entry.freq, entry.vector_index))
                i = i + 1
            return srsly.json_dumps(tuples)

        def serialize_aliases():
            i = 1
            headers = []
            indices_lists = []
            probs_lists = []
            for alias_hash, alias_index in sorted(self._alias_index.items(), key=lambda x: x[1]):
                alias = self._aliases_table[alias_index]
                assert alias_index == i
                candidate_length = len(alias.entry_indices)
                headers.append((alias_hash, candidate_length))
                indices_lists.append(alias.entry_indices)
                probs_lists.append(alias.probs)
                i = i + 1
            headers_dump = srsly.json_dumps(headers)
            indices_dump = srsly.json_dumps(indices_lists)
            probs_dump = srsly.json_dumps(probs_lists)
            return srsly.json_dumps((headers_dump, indices_dump, probs_dump))

        serializers = {
            "header": serialize_header,
            "entity_vectors": lambda: srsly.json_dumps(self._vectors_table),
            "entries": serialize_entries,
            "aliases": serialize_aliases,
        }
        return util.to_bytes(serializers, [])

    def from_bytes(self, bytes_data, *, exclude=tuple()):
        """Load state from a binary string.
        """
        def deserialize_header(b):
            header = srsly.json_loads(b)
            nr_entities = header[0]
            nr_aliases = header[1]
            entity_vector_length = header[2]
            self._initialize_entities(nr_entities)
            self._initialize_vectors(nr_entities)
            self._initialize_aliases(nr_aliases)
            self.entity_vector_length = entity_vector_length

        def deserialize_vectors(b):
            self._vectors_table = srsly.json_loads(b)

        def deserialize_entries(b):
            cdef KBEntryC entry
            tuples = srsly.json_loads(b)
            i = 1
            for (entity_hash, freq, vector_index) in tuples:
                entry.entity_hash = entity_hash
                entry.freq = freq
                entry.vector_index = vector_index
                entry.feats_row = -1  # Features table currently not implemented
                self._entries[i] = entry
                self._entry_index[entity_hash] = i
                i += 1

        def deserialize_aliases(b):
            cdef AliasC alias
            i = 1
            all_data = srsly.json_loads(b)
            headers = srsly.json_loads(all_data[0])
            indices = srsly.json_loads(all_data[1])
            probs = srsly.json_loads(all_data[2])
            for header, indices, probs in zip(headers, indices, probs):
                alias_hash, candidate_length = header
                alias.entry_indices = indices
                alias.probs = probs
                self._aliases_table[i] = alias
                self._alias_index[alias_hash] = i
                i += 1

        setters = {
            "header": deserialize_header,
            "entity_vectors": deserialize_vectors,
            "entries": deserialize_entries,
            "aliases": deserialize_aliases,
        }
        util.from_bytes(bytes_data, setters, exclude)
        return self

    def to_disk(self, path, exclude: Iterable[str] = SimpleFrozenList()):
        path = ensure_path(path)
        if not path.exists():
            path.mkdir(parents=True)
        if not path.is_dir():
            raise ValueError(Errors.E928.format(loc=path))
        serialize = {}
        serialize["contents"] = lambda p: self.write_contents(p)
        serialize["strings.json"] = lambda p: self.vocab.strings.to_disk(p)
        util.to_disk(path, serialize, exclude)

    def from_disk(self, path, exclude: Iterable[str] = SimpleFrozenList()):
        path = ensure_path(path)
        if not path.exists():
            raise ValueError(Errors.E929.format(loc=path))
        if not path.is_dir():
            raise ValueError(Errors.E928.format(loc=path))
        deserialize: Dict[str, Callable[[Any], Any]] = {}
        deserialize["contents"] = lambda p: self.read_contents(p)
        deserialize["strings.json"] = lambda p: self.vocab.strings.from_disk(p)
        util.from_disk(path, deserialize, exclude)

    def write_contents(self, file_path):
        cdef Writer writer = Writer(file_path)
        writer.write_header(self.get_size_entities(), self.entity_vector_length)

        # dumping the entity vectors in their original order
        i = 0
        for entity_vector in self._vectors_table:
            for element in entity_vector:
                writer.write_vector_element(element)
            i = i+1

        # dumping the entry records in the order in which they are in the _entries vector.
        # index 0 is a dummy object not stored in the _entry_index and can be ignored.
        i = 1
        for entry_hash, entry_index in sorted(self._entry_index.items(), key=lambda x: x[1]):
            entry = self._entries[entry_index]
            assert entry.entity_hash == entry_hash
            assert entry_index == i
            writer.write_entry(entry.entity_hash, entry.freq, entry.vector_index)
            i = i+1

        writer.write_alias_length(self.get_size_aliases())

        # dumping the aliases in the order in which they are in the _alias_index vector.
        # index 0 is a dummy object not stored in the _aliases_table and can be ignored.
        i = 1
        for alias_hash, alias_index in sorted(self._alias_index.items(), key=lambda x: x[1]):
            alias = self._aliases_table[alias_index]
            assert alias_index == i

            candidate_length = len(alias.entry_indices)
            writer.write_alias_header(alias_hash, candidate_length)

            for j in range(0, candidate_length):
                writer.write_alias(alias.entry_indices[j], alias.probs[j])

            i = i+1

        writer.close()

    def read_contents(self, file_path):
        cdef hash_t entity_hash
        cdef hash_t alias_hash
        cdef int64_t entry_index
        cdef float freq, prob
        cdef int32_t vector_index
        cdef KBEntryC entry
        cdef AliasC alias
        cdef float vector_element

        cdef Reader reader = Reader(file_path)

        # STEP 0: load header and initialize KB
        cdef int64_t nr_entities
        cdef int64_t entity_vector_length
        reader.read_header(&nr_entities, &entity_vector_length)

        self._initialize_entities(nr_entities)
        self._initialize_vectors(nr_entities)
        self.entity_vector_length = entity_vector_length

        # STEP 1: load entity vectors
        cdef int i = 0
        cdef int j = 0
        while i < nr_entities:
            entity_vector = float_vec(entity_vector_length)
            j = 0
            while j < entity_vector_length:
                reader.read_vector_element(&vector_element)
                entity_vector[j] = vector_element
                j = j+1
            self._vectors_table[i] = entity_vector
            i = i+1

        # STEP 2: load entities
        # we assume that the entity data was written in sequence
        # index 0 is a dummy object not stored in the _entry_index and can be ignored.
        i = 1
        while i <= nr_entities:
            reader.read_entry(&entity_hash, &freq, &vector_index)

            entry.entity_hash = entity_hash
            entry.freq = freq
            entry.vector_index = vector_index
            entry.feats_row = -1    # Features table currently not implemented

            self._entries[i] = entry
            self._entry_index[entity_hash] = i

            i += 1

        # check that all entities were read in properly
        assert nr_entities == self.get_size_entities()

        # STEP 3: load aliases
        cdef int64_t nr_aliases
        reader.read_alias_length(&nr_aliases)
        self._initialize_aliases(nr_aliases)

        cdef int64_t nr_candidates
        cdef vector[int64_t] entry_indices
        cdef vector[float] probs

        i = 1
        # we assume the alias data was written in sequence
        # index 0 is a dummy object not stored in the _entry_index and can be ignored.
        while i <= nr_aliases:
            reader.read_alias_header(&alias_hash, &nr_candidates)
            entry_indices = vector[int64_t](nr_candidates)
            probs = vector[float](nr_candidates)

            for j in range(0, nr_candidates):
                reader.read_alias(&entry_index, &prob)
                entry_indices[j] = entry_index
                probs[j] = prob

            alias.entry_indices = entry_indices
            alias.probs = probs

            self._aliases_table[i] = alias
            self._alias_index[alias_hash] = i

            i += 1

        # check that all aliases were read in properly
        assert nr_aliases == self.get_size_aliases()


cdef class Writer:
    def __init__(self, path):
        assert isinstance(path, Path)
        content = bytes(path)
        cdef bytes bytes_loc = content.encode('utf8') if type(content) == str else content
        self._fp = fopen(<char*>bytes_loc, 'wb')
        if not self._fp:
            raise IOError(Errors.E146.format(path=path))
        fseek(self._fp, 0, 0)

    def close(self):
        cdef size_t status = fclose(self._fp)
        assert status == 0

    cdef int write_header(self, int64_t nr_entries, int64_t entity_vector_length) except -1:
        self._write(&nr_entries, sizeof(nr_entries))
        self._write(&entity_vector_length, sizeof(entity_vector_length))

    cdef int write_vector_element(self, float element) except -1:
        self._write(&element, sizeof(element))

    cdef int write_entry(self, hash_t entry_hash, float entry_freq, int32_t vector_index) except -1:
        self._write(&entry_hash, sizeof(entry_hash))
        self._write(&entry_freq, sizeof(entry_freq))
        self._write(&vector_index, sizeof(vector_index))
        # Features table currently not implemented and not written to file

    cdef int write_alias_length(self, int64_t alias_length) except -1:
        self._write(&alias_length, sizeof(alias_length))

    cdef int write_alias_header(self, hash_t alias_hash, int64_t candidate_length) except -1:
        self._write(&alias_hash, sizeof(alias_hash))
        self._write(&candidate_length, sizeof(candidate_length))

    cdef int write_alias(self, int64_t entry_index, float prob) except -1:
        self._write(&entry_index, sizeof(entry_index))
        self._write(&prob, sizeof(prob))

    cdef int _write(self, void* value, size_t size) except -1:
        status = fwrite(value, size, 1, self._fp)
        assert status == 1, status


cdef class Reader:
    def __init__(self, path):
        content = bytes(path)
        cdef bytes bytes_loc = content.encode('utf8') if type(content) == str else content
        self._fp = fopen(<char*>bytes_loc, 'rb')
        if not self._fp:
            PyErr_SetFromErrno(IOError)
        status = fseek(self._fp, 0, 0)  # this can be 0 if there is no header

    def __dealloc__(self):
        fclose(self._fp)

    cdef int read_header(self, int64_t* nr_entries, int64_t* entity_vector_length) except -1:
        status = self._read(nr_entries, sizeof(int64_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="header"))

        status = self._read(entity_vector_length, sizeof(int64_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="vector length"))

    cdef int read_vector_element(self, float* element) except -1:
        status = self._read(element, sizeof(float))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="vector element"))

    cdef int read_entry(self, hash_t* entity_hash, float* freq, int32_t* vector_index) except -1:
        status = self._read(entity_hash, sizeof(hash_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="entity hash"))

        status = self._read(freq, sizeof(float))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="entity freq"))

        status = self._read(vector_index, sizeof(int32_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="vector index"))

        if feof(self._fp):
            return 0
        else:
            return 1

    cdef int read_alias_length(self, int64_t* alias_length) except -1:
        status = self._read(alias_length, sizeof(int64_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="alias length"))

    cdef int read_alias_header(self, hash_t* alias_hash, int64_t* candidate_length) except -1:
        status = self._read(alias_hash, sizeof(hash_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="alias hash"))

        status = self._read(candidate_length, sizeof(int64_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="candidate length"))

    cdef int read_alias(self, int64_t* entry_index, float* prob) except -1:
        status = self._read(entry_index, sizeof(int64_t))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="entry index"))

        status = self._read(prob, sizeof(float))
        if status < 1:
            if feof(self._fp):
                return 0  # end of file
            raise IOError(Errors.E145.format(param="prior probability"))

    cdef int _read(self, void* value, size_t size) except -1:
        status = fread(value, size, 1, self._fp)
        return status