Skip to content

edsnlp.pipelines.misc.sections

patterns

These section titles were extracted from a work performed by Ivan Lerner at AP-HP. It supplied a number of documents annotated for section titles.

The section titles were reviewed by Gilles Chatellier, who gave meaningful insights.

See sections/section-dataset notebook for detail.

sections

Sections

Bases: GenericMatcher

Divides the document into sections.

By default, we are using a dataset of documents annotated for section titles, using the work done by Ivan Lerner, reviewed by Gilles Chatellier.

Detected sections are :

  • allergies ;
  • antécédents ;
  • antécédents familiaux ;
  • traitements entrée ;
  • conclusion ;
  • conclusion entrée ;
  • habitus ;
  • correspondants ;
  • diagnostic ;
  • données biométriques entrée ;
  • examens ;
  • examens complémentaires ;
  • facteurs de risques ;
  • histoire de la maladie ;
  • actes ;
  • motif ;
  • prescriptions ;
  • traitements sortie.

The component looks for section titles within the document, and stores them in the section_title extension.

For ease-of-use, the component also populates a section extension, which contains a list of spans corresponding to the "sections" of the document. These span from the start of one section title to the next, which can introduce obvious bias should an intermediate section title goes undetected.

PARAMETER DESCRIPTION
nlp

spaCy pipeline object.

TYPE: Language

sections

Dictionary of terms to look for.

TYPE: Dict[str, List[str]]

attr

Default attribute to match on.

TYPE: str

ignore_excluded

Whether to skip excluded tokens.

TYPE: bool

Source code in edsnlp/pipelines/misc/sections/sections.py 
 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
class Sections(GenericMatcher):
    """
    Divides the document into sections.

    By default, we are using a dataset of documents annotated for section titles,
    using the work done by Ivan Lerner, reviewed by Gilles Chatellier.

    Detected sections are :

    - allergies ;
    - antécédents ;
    - antécédents familiaux ;
    - traitements entrée ;
    - conclusion ;
    - conclusion entrée ;
    - habitus ;
    - correspondants ;
    - diagnostic ;
    - données biométriques entrée ;
    - examens ;
    - examens complémentaires ;
    - facteurs de risques ;
    - histoire de la maladie ;
    - actes ;
    - motif ;
    - prescriptions ;
    - traitements sortie.

    The component looks for section titles within the document,
    and stores them in the `section_title` extension.

    For ease-of-use, the component also populates a `section` extension,
    which contains a list of spans corresponding to the "sections" of the
    document. These span from the start of one section title to the next,
    which can introduce obvious bias should an intermediate section title
    goes undetected.

    Parameters
    ----------
    nlp : Language
        spaCy pipeline object.
    sections : Dict[str, List[str]]
        Dictionary of terms to look for.
    attr : str
        Default attribute to match on.
    ignore_excluded : bool
        Whether to skip excluded tokens.
    """

    def __init__(
        self,
        nlp: Language,
        sections: Dict[str, List[str]],
        add_patterns: bool,
        attr: str,
        ignore_excluded: bool,
    ):

        logger.warning(
            "The component Sections is still in Beta. Use at your own risks."
        )

        if sections is None:
            sections = patterns.sections

        self.add_patterns = add_patterns
        if add_patterns:
            for k, v in sections.items():
                sections[k] = [r"\n[^\n]{0,5}" + ent + r"[^\n]{0,5}\n" for ent in v]

        super().__init__(
            nlp,
            terms=None,
            regex=sections,
            attr=attr,
            ignore_excluded=ignore_excluded,
        )

        self.set_extensions()

        if not nlp.has_pipe("normalizer") and not not nlp.has_pipe("eds.normalizer"):
            logger.warning("You should add pipe `eds.normalizer`")

    @staticmethod
    def set_extensions():

        if not Span.has_extension("section_title"):
            Span.set_extension("section_title", default=None)

        if not Span.has_extension("section"):
            Span.set_extension("section", default=None)

    # noinspection PyProtectedMember
    def __call__(self, doc: Doc) -> Doc:
        """
        Divides the doc into sections

        Parameters
        ----------
        doc:
            spaCy Doc object

        Returns
        -------
        doc:
            spaCy Doc object, annotated for sections
        """
        titles = filter_spans(self.process(doc))

        if self.add_patterns:
            # Remove preceding newline
            titles = [
                Span(doc, title.start + 1, title.end - 1, label=title.label_)
                for title in titles
            ]

        sections = []

        for t1, t2 in zip(titles[:-1], titles[1:]):
            section = Span(doc, t1.start, t2.start, label=t1.label)
            section._.section_title = t1
            sections.append(section)

        if titles:
            t = titles[-1]
            section = Span(doc, t.start, len(doc), label=t.label)
            section._.section_title = t
            sections.append(section)

        doc.spans["sections"] = sections
        doc.spans["section_titles"] = titles

        return doc
__call__(doc)

Divides the doc into sections

PARAMETER DESCRIPTION
doc

spaCy Doc object

TYPE: Doc

RETURNS DESCRIPTION
doc

spaCy Doc object, annotated for sections
Source code in edsnlp/pipelines/misc/sections/sections.py 
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
def __call__(self, doc: Doc) -> Doc:
    """
    Divides the doc into sections

    Parameters
    ----------
    doc:
        spaCy Doc object

    Returns
    -------
    doc:
        spaCy Doc object, annotated for sections
    """
    titles = filter_spans(self.process(doc))

    if self.add_patterns:
        # Remove preceding newline
        titles = [
            Span(doc, title.start + 1, title.end - 1, label=title.label_)
            for title in titles
        ]

    sections = []

    for t1, t2 in zip(titles[:-1], titles[1:]):
        section = Span(doc, t1.start, t2.start, label=t1.label)
        section._.section_title = t1
        sections.append(section)

    if titles:
        t = titles[-1]
        section = Span(doc, t.start, len(doc), label=t.label)
        section._.section_title = t
        sections.append(section)

    doc.spans["sections"] = sections
    doc.spans["section_titles"] = titles

    return doc
Back to top