Skip to content

Extraction Module

TripletExtractor

TripletExtractor

Bases: ABC

Abstract base class for triplet extraction algorithms.

Triplets are instantiated as Triplet objects that consist of SpanAnnotation objects.

Thus, to create a Triplet, you create the

Source code in narrativegraphs/nlp/triplets/common.py 
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
class TripletExtractor(ABC):
    """
    Abstract base class for triplet extraction algorithms.

    Triplets are instantiated as Triplet objects that consist of SpanAnnotation objects.

    Thus, to create a Triplet, you create the
    """

    @abstractmethod
    def extract(self, text: str) -> list[Triplet]:
        """Single document extraction
        Args:
            text: a raw text string

        Returns:
            extracted triplets
        """
        pass

    def batch_extract(
        self, texts: Iterable[str], n_cpu: int = 1, **kwargs
    ) -> Generator[list[Triplet], None, None]:
        """Multiple-document extraction
        Args:
            texts: an iterable of raw text strings; may be a generator, so be mindful
                of consuming items
            n_cpu: number of CPUs to use
            **kwargs: other keyword arguments for your own class

        Returns:
            should yield triplets per text in the same order as texts iterable

        """
        for text in texts:
            yield self.extract(text)

extract(text) abstractmethod

Single document extraction Args: text: a raw text string

Returns:

Type Description
list[Triplet]

extracted triplets
Source code in narrativegraphs/nlp/triplets/common.py 
25
26
27
28
29
30
31
32
33
34
@abstractmethod
def extract(self, text: str) -> list[Triplet]:
    """Single document extraction
    Args:
        text: a raw text string

    Returns:
        extracted triplets
    """
    pass

batch_extract(texts, n_cpu=1, **kwargs)

Multiple-document extraction Args: texts: an iterable of raw text strings; may be a generator, so be mindful of consuming items n_cpu: number of CPUs to use **kwargs: other keyword arguments for your own class

Returns:

Type Description
None

should yield triplets per text in the same order as texts iterable
Source code in narrativegraphs/nlp/triplets/common.py 
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
def batch_extract(
    self, texts: Iterable[str], n_cpu: int = 1, **kwargs
) -> Generator[list[Triplet], None, None]:
    """Multiple-document extraction
    Args:
        texts: an iterable of raw text strings; may be a generator, so be mindful
            of consuming items
        n_cpu: number of CPUs to use
        **kwargs: other keyword arguments for your own class

    Returns:
        should yield triplets per text in the same order as texts iterable

    """
    for text in texts:
        yield self.extract(text)

SpacyTripletExtractor

Bases: TripletExtractor

Base class for implementing triplet extraction based on spaCy docs.

Override extract_triplets_from_sent for extracting triplets sentence by sentence.

Override extract_triplets_from_doc for extracting with the full Doc context.

The SpanAnnotation objects of Triplet objects can conveniently be created from a spaCy Span object with SpanAnnotation.from_span().

Source code in narrativegraphs/nlp/triplets/spacy/common.py 
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
class SpacyTripletExtractor(TripletExtractor):
    """Base class for implementing triplet extraction based on spaCy docs.

    Override `extract_triplets_from_sent` for extracting triplets sentence by sentence.

    Override `extract_triplets_from_doc` for extracting with the full Doc context.

    The `SpanAnnotation` objects of `Triplet` objects can conveniently be created from
    a spaCy `Span` object with `SpanAnnotation.from_span()`.
    """

    def __init__(
        self, model_name: str = None, split_sentence_on_double_line_break: bool = True
    ):
        """
        Args:
            model_name: name of the spaCy model to use
            split_sentence_on_double_line_break: adds extra sentence boundaries on
                double line breaks ("\n\n")
        """
        if model_name is None:
            model_name = "en_core_web_sm"
        self.nlp = ensure_spacy_model(model_name)
        if split_sentence_on_double_line_break:
            self.nlp.add_pipe("custom_sentencizer", before="parser")

    @abstractmethod
    def extract_triplets_from_sent(self, sent: Span) -> list[Triplet]:
        """Extract triplets from a SpaCy sentence.
        Args:
            sent: A SpaCy Span object representing the whole sentence

        Returns:
            extracted triplets
        """
        pass

    def extract_triplets_from_doc(self, doc: Doc) -> list[Triplet]:
        """Extract triplets from a Doc
        Args:
            doc: A SpaCy Doc object

        Returns:
            extracted triplets
        """
        triplets = []
        for sent in doc.sents:
            sent_triplets = self.extract_triplets_from_sent(sent)
            if sent_triplets is not None:
                triplets.extend(sent_triplets)
        return triplets

    def extract(self, text: str) -> list[Triplet]:
        text = self.nlp(text)
        return self.extract_triplets_from_doc(text)

    def batch_extract(
        self, texts: list[str], n_cpu: int = 1, batch_size: int = None
    ) -> Generator[list[Triplet], None, None]:
        if batch_size is None:
            batch_size = calculate_batch_size(texts, n_cpu)
        if n_cpu > 1:
            _logger.info(
                "Using multiple CPU cores.Progress bars may stand still at first."
            )
        for doc in self.nlp.pipe(texts, n_process=n_cpu, batch_size=batch_size):
            yield self.extract_triplets_from_doc(doc)

__init__(model_name=None, split_sentence_on_double_line_break=True) 

    Args:
        model_name: name of the spaCy model to use
        split_sentence_on_double_line_break: adds extra sentence boundaries on
            double line breaks ("

")

Source code in narrativegraphs/nlp/triplets/spacy/common.py 
27
28
29
30
31
32
33
34
35
36
37
38
39
40
def __init__(
    self, model_name: str = None, split_sentence_on_double_line_break: bool = True
):
    """
    Args:
        model_name: name of the spaCy model to use
        split_sentence_on_double_line_break: adds extra sentence boundaries on
            double line breaks ("\n\n")
    """
    if model_name is None:
        model_name = "en_core_web_sm"
    self.nlp = ensure_spacy_model(model_name)
    if split_sentence_on_double_line_break:
        self.nlp.add_pipe("custom_sentencizer", before="parser")

extract_triplets_from_sent(sent) abstractmethod

Extract triplets from a SpaCy sentence. Args: sent: A SpaCy Span object representing the whole sentence

Returns:

Type Description
list[Triplet]

extracted triplets
Source code in narrativegraphs/nlp/triplets/spacy/common.py 
42
43
44
45
46
47
48
49
50
51
@abstractmethod
def extract_triplets_from_sent(self, sent: Span) -> list[Triplet]:
    """Extract triplets from a SpaCy sentence.
    Args:
        sent: A SpaCy Span object representing the whole sentence

    Returns:
        extracted triplets
    """
    pass

extract_triplets_from_doc(doc)

Extract triplets from a Doc Args: doc: A SpaCy Doc object

Returns:

Type Description
list[Triplet]

extracted triplets
Source code in narrativegraphs/nlp/triplets/spacy/common.py 
53
54
55
56
57
58
59
60
61
62
63
64
65
66
def extract_triplets_from_doc(self, doc: Doc) -> list[Triplet]:
    """Extract triplets from a Doc
    Args:
        doc: A SpaCy Doc object

    Returns:
        extracted triplets
    """
    triplets = []
    for sent in doc.sents:
        sent_triplets = self.extract_triplets_from_sent(sent)
        if sent_triplets is not None:
            triplets.extend(sent_triplets)
    return triplets