spaCy/website/docs/usage/visualizers.jade

//- 💫 DOCS > USAGE > VISUALIZERS

include ../../_includes/_mixins

p
    |  As of v2.0, our popular visualizers, #[+a(DEMOS_URL + "/displacy") displaCy]
    |  and #[+a(DEMOS_URL + "/displacy-ent") displaCy #[sup ENT]] are finally an
    |  official part of the library. Visualizing a dependency parse or named
    |  entities in a text is not only a fun NLP demo – it can also be incredibly
    |  helpful in speeding up development and debugging your code and training
    |  process. Instead of printing a list of dependency labels or entity spans,
    |  you can simply pass your #[code Doc] objects to #[code displacy] and view
    |  the visualizations in your browser, or export them as HTML files or
    |  vector graphics.

p
    |  If you're running a #[+a("https://jupyter.org") Jupyter] notebook,
    |  displaCy will detect this and return the markup in a format
    |  #[+a("#jupyter") ready to be rendered and exported].

+aside("What about the old visualizers?")
    |  Our JavaScript-based visualizers #[+src(gh("displacy")) displacy.js] and
    |  #[+src(gh("displacy-ent")) displacy-ent.js] will still be available on
    |  GitHub. If you're looking to implement web-based visualizations, we
    |  generally recommend using those instead of spaCy's built-in
    |  #[code displacy] module. It'll allow your application to perform all
    |  rendering on the client and only rely on the server for the text
    |  processing. The generated markup is also more compatible with modern web
    |  standards.

+h(2, "getting-started") Getting started
    +tag-new(2)

p
    |  The quickest way visualize  #[code Doc] is to use
    |  #[+api("displacy#serve") #[code displacy.serve]]. This will spin up a
    |  simple web server and let you view the result straight from your browser.
    |  displaCy can either take a single #[code Doc] or a list of #[code Doc]
    |  objects as its first argument. This lets you construct them however you
    |  like – using any model or modifications you like.

+h(3, "dep") Visualizing the dependency parse

p
    |  The dependency visualizer, #[code dep], shows part-of-speech tags
    |  and syntactic dependencies.

+code("Dependency example").
    import spacy
    from spacy import displacy

    nlp = spacy.load('en')
    doc = nlp(u'This is a sentence.')
    displacy.serve(doc, style='dep')

+codepen("f0e85b64d469d6617251d8241716d55f", 370)

p
    |  The argument #[code options] lets you specify a dictionary of settings
    |  to customise the layout, for example:

+aside("Important note")
    |  There's currently a known issue with the #[code compact] mode for
    |  sentences with short arrows and long dependency labels, that causes labels
    |  longer than the arrow to wrap. So if you come across this problem,
    |  especially when using custom labels, you'll have to increase the
    |  #[code distance] setting in the #[code options] to allow longer arcs.

+table(["Name", "Type", "Description", "Default"])
    +row
        +cell #[code compact]
        +cell bool
        +cell "Compact mode" with square arrows that takes up less space.
        +cell #[code False]

    +row
        +cell #[code color]
        +cell unicode
        +cell Text color (HEX, RGB or color names).
        +cell #[code '#000000']

    +row
        +cell #[code bg]
        +cell unicode
        +cell Background color (HEX, RGB or color names).
        +cell #[code '#ffffff']

    +row
        +cell #[code font]
        +cell unicode
        +cell Font name or font family for all text.
        +cell #[code 'Arial']

p
    |  For a list of all available options, see the
    |  #[+api("displacy#options") #[code displacy] API documentation].

+aside-code("Options example").
    options = {'compact': True, 'bg': '#09a3d5',
               'color': 'white', 'font': 'Source Sans Pro'}
    displacy.serve(doc, style='dep', options=options)

+codepen("39c02c893a84794353de77a605d817fd", 360)

+h(3, "ent") Visualizing the entity recognizer

p
    |  The entity visualizer, #[code ent], highlights named entities and
    |  their labels in a text.

+code("Named Entity example").
    import spacy
    from spacy import displacy

    text = """But Google is starting from behind. The company made a late push
    into hardware, and Apple’s Siri, available on iPhones, and Amazon’s Alexa
    software, which runs on its Echo and Dot devices, have clear leads in
    consumer adoption."""

    nlp = spacy.load('custom_ner_model')
    doc = nlp(text)
    displacy.serve(doc, style='ent')

+codepen("a73f8b68f9af3157855962b283b364e4", 345)

p The entity visualizer lets you customise the following #[code options]:

+table(["Name", "Type", "Description", "Default"])
    +row
        +cell #[code ents]
        +cell list
        +cell
            |  Entity types to highlight (#[code None] for all types).
        +cell #[code None]

    +row
        +cell #[code colors]
        +cell dict
        +cell
            |  Color overrides. Entity types in lowercase should be mapped to
            |  color names or values.
        +cell #[code {}]

p
    |  If you specify a list of #[code ents], only those entity types will be
    |  rendered – for example, you can choose to display #[code PERSON] entities.
    |  Internally, the visualizer knows nothing about available entity types and
    |  will render whichever spans and labels it receives. This makes it
    |  especially easy to work with custom entity types. By default, displaCy
    |  comes with colours for all
    |  #[+a("/docs/api/annotation#named-entities") entity types supported by spaCy].
    |  If you're using custom entity types, you can use the #[code colors]
    |  setting to add your own colours for them.

+aside-code("Options example").
    colors = {'ORG': 'linear-gradient(90deg, #aa9cfc, #fc9ce7)'}
    options = {'ents': ['ORG'], 'colors': colors}
    displacy.serve(doc, style='ent', options=options)

+codepen("f42ec690762b6f007022a7acd6d0c7d4", 300)

p
    |  The above example uses a little trick: Since the background colour values
    |  are added as the #[code background] style attribute, you can use any
    |  #[+a("https://tympanus.net/codrops/css_reference/background/") valid background value]
    |  or shorthand — including gradients and even images!

+h(3, "ent-titles") Adding titles to documents

p
    |  Rendering several large documents on one page can easily become confusing.
    |  To add a headline to each visualization, you can add a #[code title] to
    |  its #[code user_data]. User data is never touched or modified by spaCy.

+code.
    doc = nlp(u'This is a sentence about Google.')
    doc.user_data['title'] = 'This is a title'
    displacy.serve(doc, style='ent')

p
    |  This feature is espeically handy if you're using displaCy to compare
    |  performance at different stages of a process, e.g. during training. Here
    |  you could use the title for a brief description of the text example and
    |  the number of iterations.

+h(2, "render") Rendering visualizations

p
    |  If you don't need the web server and just want to generate the markup
    |  – for example, to export it to a file or serve it in a custom
    |  way – you can use #[+api("displacy#render") #[code displacy.render]].
    |  It works the same way, but returns a string containing the markup.

+code("Example").
    import spacy
    from spacy import displacy

    nlp = spacy.load('en')
    doc1 = nlp(u'This is a sentence.')
    doc2 = nlp(u'This is another sentence.')
    html = displacy.render([doc1, doc2], style='dep', page=True)

p
    |  #[code page=True] renders the markup wrapped as a full HTML page.
    |  For minified and more compact HTML markup, you can set #[code minify=True].
    |  If you're rendering a dependency parse, you can also export it as an
    |  #[code .svg] file.

+aside("What's SVG?")
    |  Unlike other image formats, the SVG (Scalable Vector Graphics) uses XML
    |  markup that's easy to manipulate
    |  #[+a("https://www.smashingmagazine.com/2014/11/styling-and-animating-svgs-with-css/") using CSS] or
    |  #[+a("https://css-tricks.com/smil-is-dead-long-live-smil-a-guide-to-alternatives-to-smil-features/") JavaScript].
    |  Essentially, SVG lets you design with code, which makes it a perfect fit
    |  for visualizing dependency trees. SVGs can be embedded online in an
    |  #[code <img>] tag, or inlined in an HTML document. They're also
    |  pretty easy to #[+a("https://convertio.co/image-converter/") convert].

+code.
    svg = displacy.render(doc, style='dep')
    output_path = Path('/images/sentence.svg')
    output_path.open('w', encoding='utf-8').write(svg)

+infobox("Important note")
    |  Since each visualization is generated as a separate SVG, exporting
    |  #[code .svg] files only works if you're rendering #[strong one single doc]
    |  at a time. (This makes sense – after all, each visualization should be
    |  a standalone graphic.) So instead of rendering all #[code Doc]s at one,
    |  loop over them and export them separately.


+h(3, "examples-export-svg") Example: Export SVG graphics of dependency parses

+code("Example").
    import spacy
    from spacy import displacy
    from pathlib import Path

    nlp = spacy.load('en')
    sentences = ["This is an example.", "This is another one."]
    for sent in sentences:
        doc = nlp(sentence)
        svg = displacy.render(doc, style='dep')
        file_name = '-'.join([w.text for w in doc if not w.is_punct]) + '.svg'
        output_path = Path('/images/' + file_name)
        output_path.open('w', encoding='utf-8').write(svg)

p
    |  The above code will generate the dependency visualizations and them to
    |  two files, #[code This-is-an-example.svg] and #[code This-is-another-one.svg].


+h(2, "jupyter") Using displaCy in Jupyter notebooks

p
    |  displaCy is able to detect whether you're working in a
    |  #[+a("https://jupyter.org") Jupyter] notebook, and will return markup
    |  that can be rendered in a cell straight away. When you export your
    |  notebook, the visualizations will be included as HTML.

+code("Jupyter Example").
    # don't forget to install a model, e.g.: python -m spacy download en
    import spacy
    from spacy import displacy

    doc = nlp(u'Rats are various medium-sized, long-tailed rodents.')
    displacy.render(doc, style='dep')

    doc2 = nlp(LONG_NEWS_ARTICLE)
    displacy.render(doc2, style='ent')

+aside("Enabling or disabling Jupyter mode")
    |  To explicitly enable or disable "Jupyter mode", you can use the
    |  #[code jupyter] keyword argument – e.g. to return raw HTML in a notebook,
    |  or to force Jupyter rendering if auto-detection fails.

+image("/assets/img/docs/displacy_jupyter.jpg", 700, false, "Example of using the displaCy dependency and named entity visualizer in a Jupyter notebook")

p
    |  Internally, displaCy imports #[code display] and #[code HTML] from
    |  #[code IPython.core.display] and returns a Jupyter HTML object. If you
    |  were doing it manually, it'd look like this:

+code.
    from IPython.core.display import display, HTML

    html = displacy.render(doc, style='dep')
    return display(HTML(html))

+h(2, "manual-usage") Rendering data manually

p
    |  You can also use displaCy to manually render data. This can be useful if
    |  you want to visualize output from other libraries, like
    |  #[+a("http://www.nltk.org") NLTK] or
    |  #[+a("https://github.com/tensorflow/models/tree/master/syntaxnet") SyntaxNet].
    |  Simply convert the dependency parse or recognised entities to displaCy's
    |  format and set #[code manual=True] on either #[code render()] or
    |  #[code serve()].

+aside-code("Example").
    ex = [{'text': 'But Google is starting from behind.',
           'ents': [{'start': 4, 'end': 10, 'label': 'ORG'}],
           'title': None}]
    html = displacy.render(ex, style='ent', manual=True)

+code("DEP input").
    {
        'words': [
            {'text': 'This', 'tag': 'DT'},
            {'text': 'is', 'tag': 'VBZ'},
            {'text': 'a', 'tag': 'DT'},
            {'text': 'sentence', 'tag': 'NN'}],
        'arcs': [
            {'start': 0, 'end': 1, 'label': 'nsubj', 'dir': 'left'},
            {'start': 2, 'end': 3, 'label': 'det', 'dir': 'left'},
            {'start': 1, 'end': 3, 'label': 'attr', 'dir': 'right'}]
    }

+code("ENT input").
    {
        'text': 'But Google is starting from behind.',
        'ents': [{'start': 4, 'end': 10, 'label': 'ORG'}],
        'title': None
    }

+h(2, "webapp") Using displaCy in a web application

p
    |  If you want to use the visualizers as part of a web application, for
    |  example to create something like our
    |  #[+a(DEMOS_URL + "/displacy") online demo], it's not recommended to
    |  simply wrap and serve the displaCy renderer. Instead, you should only
    |  rely on the server to perform spaCy's processing capabilities, and use
    |  #[+a(gh("displacy")) displaCy.js] to render the JSON-formatted output.

+aside("Why not return the HTML by the server?")
    |  It's certainly possible to just have your server return the markup.
    |  But outputting raw, unsanitised HTML is risky and makes your app vulnerable to
    |  #[+a("https://en.wikipedia.org/wiki/Cross-site_scripting") cross-site scripting]
    |  (XSS). All your user needs to do is find a way to make spaCy return text
    |  like #[code <script src="malicious-code.js"><script>], which
    |  is pretty easy in NER mode. Instead of relying on the server to render
    |  and sanitise HTML, you can do this on the client in JavaScript.
    |  displaCy.js creates the markup as DOM nodes and will never insert raw
    |  HTML.

p
    |  The #[code parse_deps] function takes a #[code Doc] object and returns
    |  a dictionary in a format that can be rendered by displaCy.

+code("Example").
    import spacy
    from spacy import displacy

    nlp = spacy.load('en')

    def displacy_service(text):
        doc = nlp(text)
        return displacy.parse_deps(doc)

p
    |  Using a library like #[+a("https://falconframework.org/") Falcon] or
    |  #[+a("http://www.hug.rest/") Hug], you can easily turn the above code
    |  into a simple REST API that receives a text and returns a JSON-formatted
    |  parse. In your front-end, include #[+a(gh("displacy")) displacy.js] and
    |  initialise it with the API URL and the ID or query selector of the
    |  container to render the visualisation in, e.g. #[code '#displacy'] for
    |  #[code <div id="displacy">].

+code("script.js", "javascript").
    var displacy = new displaCy('http://localhost:8080', {
        container: '#displacy'
    })

    function parse(text) {
        displacy.parse(text);
    }

p
    |  When you call #[code parse()], it will make a request to your API,
    |  receive the JSON-formatted parse and render it in your container. To
    |  create an interactive experience, you could trigger this function by
    |  a button and read the text from an #[code <input>] field.