from turtle import down import spacy from spacy import displacy import random from spacy.tokens import Span import gradio as gr import pandas as pd # import cairosvg import base64 DEFAULT_MODEL = "en_core_web" DEFAULT_TEXT = "Apple is looking at buying U.K. startup for $1 billion." DEFAULT_TOK_ATTR = ['idx', 'text', 'pos_', 'lemma_', 'shape_', 'dep_'] DEFAULT_ENTS = ['CARDINAL', 'DATE', 'EVENT', 'FAC', 'GPE', 'LANGUAGE', 'LAW', 'LOC', 'MONEY', 'NORP', 'ORDINAL', 'ORG', 'PERCENT', 'PERSON', 'PRODUCT', 'QUANTITY', 'TIME', 'WORK_OF_ART'] DEFAULT_COLOR = "linear-gradient(90deg, #FFCA74, #7AECEC)" texts = {"en": DEFAULT_TEXT, "ca": "Apple està buscant comprar una startup del Regne Unit per mil milions de dòlars", "da": "Apple overvejer at købe et britisk startup for 1 milliard dollar.", "de": "Die ganze Stadt ist ein Startup: Shenzhen ist das Silicon Valley für Hardware-Firmen", "el": "Η άνιση κατανομή του πλούτου και του εισοδήματος, η οποία έχει λάβει τρομερές διαστάσεις, δεν δείχνει τάσεις βελτίωσης.", "es": "Apple está buscando comprar una startup del Reino Unido por mil millones de dólares.", "fi": "Itseajavat autot siirtävät vakuutusvastuun autojen valmistajille", "fr": "Apple cherche à acheter une start-up anglaise pour 1 milliard de dollars", "it": "Apple vuole comprare una startup del Regno Unito per un miliardo di dollari", "ja": "アップルがイギリスの新興企業を１０億ドルで購入を検討", "ko": "애플이 영국의 스타트업을 10억 달러에 인수하는 것을 알아보고 있다.", "lt": "Jaunikis pirmąją vestuvinę naktį iškeitė į areštinės gultą", "nb": "Apple vurderer å kjøpe britisk oppstartfirma for en milliard dollar.", "nl": "Apple overweegt om voor 1 miljard een U.K. startup te kopen", "pl": "Poczuł przyjemną woń mocnej kawy.", "pt": "Apple está querendo comprar uma startup do Reino Unido por 100 milhões de dólares", "ro": "Apple plănuiește să cumpere o companie britanică pentru un miliard de dolari", "ru": "Apple рассматривает возможность покупки стартапа из Соединённого Королевства за $1 млрд", "sv": "Apple överväger att köpa brittisk startup för 1 miljard dollar.", "zh": "作为语言而言，为世界使用人数最多的语言，目前世界有五分之一人口做为母语。"} button_css = "float: right; --tw-border-opacity: 1; border-color: rgb(229 231 235 / var(--tw-border-opacity)); --tw-gradient-from: rgb(243 244 246 / 0.7); --tw-gradient-stops: var(--tw-gradient-from), var(--tw-gradient-to, rgb(243 244 246 / 0)); --tw-gradient-to: rgb(229 231 235 / 0.8); --tw-text-opacity: 1; color: rgb(55 65 81 / var(--tw-text-opacity)); border-width: 1px; --tw-bg-opacity: 1; background-color: rgb(255 255 255 / var(--tw-bg-opacity)); background-image: linear-gradient(to bottom right, var(--tw-gradient-stops)); display: inline-flex; flex: 1 1 0%; align-items: center; justify-content: center; --tw-shadow: 0 1px 2px 0 rgb(0 0 0 / 0.05); --tw-shadow-colored: 0 1px 2px 0 var(--tw-shadow-color); box-shadow: var(--tw-ring-offset-shadow, 0 0 #0000), var(--tw-ring-shadow, 0 0 #0000), var(--tw-shadow); -webkit-appearance: button; border-radius: 0.5rem; padding-top: 0.5rem; padding-bottom: 0.5rem; padding-left: 1rem; padding-right: 1rem; font-size: 1rem; line-height: 1.5rem; font-weight: 600;" def get_all_models(): with open("requirements.txt") as f: content = f.readlines() models = [] for line in content: if "huggingface.co" in line: model = "_".join(line.split("/")[4].split("_")[:3]) if model not in models: models.append(model) return models models = get_all_models() def download_svg(svg): encode = base64.b64encode(bytes(svg, 'utf-8')) img = 'data:image/svg+xml;base64,' + str(encode)[2:-1] html = f'Download as SVG' return html # def download_png(svg): # encode = base64.b64encode(bytes(svg, 'utf-8')) # svg_uri = 'data:image/svg+xml;base64,' + str(encode)[2:-1] # output = cairosvg.svg2png(url=svg_uri) # encoded = base64.b64encode(output) # img = 'data:image/png;base64,' + str(encoded)[2:-1] # html = f'Download as PNG' # return html # def download(type, svg): # if type == 'png': # return download_png(svg) # elif type == 'svg': # return download_svg(svg) def dependency(text, col_punct, col_phrase, compact, bg, font, model): nlp = spacy.load(model + "_sm") doc = nlp(text) options = {"compact": compact, "collapse_phrases": col_phrase, "collapse_punct": col_punct, "bg": bg, "color": font} svg = displacy.render(doc, style="dep", options=options) download = download_svg(svg) return svg, download def entity(text, ents, model): nlp = spacy.load(model + "_sm") doc = nlp(text) options = {"ents": ents} svg = displacy.render(doc, style="ent", options=options) # download = download_svg('' + svg + "") return svg def token(text, attributes, model): nlp = spacy.load(model + "_sm") data = [] doc = nlp(text) for tok in doc: tok_data = [] for attr in attributes: tok_data.append(getattr(tok, attr)) data.append(tok_data) data = pd.DataFrame(data, columns=attributes) return data def default_token(text, attributes, model): nlp = spacy.load(model + "_sm") data = [] doc = nlp(text) for tok in doc: tok_data = [] for attr in attributes: tok_data.append(getattr(tok, attr)) data.append(tok_data) return data def random_vectors(text, model): nlp = spacy.load(model + "_md") doc = nlp(text) n_chunks = [chunk for chunk in doc.noun_chunks if doc.noun_chunks] words = [tok for tok in doc if not tok.is_stop and tok.pos_ not in [ 'PUNCT', "PROPN"]] str_list = n_chunks + words choice = random.choices(str_list, k=2) return round(choice[0].similarity(choice[1]), 2), choice[0].text, choice[1].text def vectors(input1, input2, model): nlp = spacy.load(model + "_md") return round(nlp(input1).similarity(nlp(input2)), 2) def span(text, span1, span2, label1, label2, model): nlp = spacy.load(model + "_sm") doc = nlp(text) if span1: idx1_1 = 0 idx1_2 = 0 idx2_1 = 0 idx2_2 = 0 span1 = [split for split in span1.split(" ") if split] span2 = [split for split in span2.split(" ") if split] for i in range(len(list(doc))): tok = list(doc)[i] if span1[0] == tok.text: idx1_1 = i if span1[-1] == tok.text: idx1_2 = i + 1 if span2[0] == tok.text: idx2_1 = i if span2[-1] == tok.text: idx2_2 = i + 1 doc.spans["sc"] = [ Span(doc, idx1_1, idx1_2, label1), Span(doc, idx2_1, idx2_2, label2), ] else: idx1_1 = 0 idx1_2 = round(len(list(doc)) / 2) idx2_1 = 0 idx2_2 = 1 doc.spans["sc"] = [ Span(doc, idx1_1, idx1_2, label1), Span(doc, idx2_1, idx2_2, label2), ] svg = displacy.render(doc, style="span") # download = download_svg(svg) return svg def get_text(model): for i in range(len(models)): model = model.split("_")[0] new_text = texts[model] return new_text demo = gr.Blocks(css="scrollbar.css") with demo: with gr.Box(): with gr.Row(): with gr.Column(): gr.Markdown("# Pipeline Visualizer") gr.Markdown( "### Visualize parts of the spaCy pipeline in an interactive demo.") with gr.Column(): gr.Image("pipeline.svg") with gr.Box(): with gr.Column(): gr.Markdown(" ## Choose a language model and text") with gr.Row(): with gr.Column(): model_input = gr.Dropdown( choices=models, value=DEFAULT_MODEL, interactive=True, label="Pretrained Pipelines") with gr.Column(): gr.Markdown("") with gr.Column(): gr.Markdown("") with gr.Column(): gr.Markdown("") with gr.Row(): with gr.Column(): text_input = gr.Textbox( value=DEFAULT_TEXT, interactive=True, label="Input Text") with gr.Column(): gr.Markdown("") button = gr.Button("Generate", variant="primary") with gr.Box(): with gr.Column(): gr.Markdown( "## [Dependency Parser](https://spacy.io/usage/visualizers#dep)") gr.Markdown( "The dependency visualizer shows part-of-speech tags and syntactic dependencies.") with gr.Row(): with gr.Column(): col_punct = gr.Checkbox( label="Collapse Punctuation", value=True) col_phrase = gr.Checkbox( label="Collapse Phrases", value=True) compact = gr.Checkbox(label="Compact", value=False) with gr.Column(): bg = gr.Textbox( label="Background Color", value=DEFAULT_COLOR) with gr.Column(): text = gr.Textbox( label="Text Color", value="black") depen_output = gr.HTML(value=dependency( DEFAULT_TEXT, True, True, False, DEFAULT_COLOR, "black", DEFAULT_MODEL)[0]) with gr.Row(): with gr.Column(): dep_button = gr.Button("Generate Dependency Parser", variant="primary") with gr.Column(): dep_download_button = gr.HTML(value=download_svg(depen_output.value)) gr.Markdown(" ") with gr.Box(): with gr.Column(): gr.Markdown( "## [Entity Recognizer](https://spacy.io/usage/visualizers#ent)") gr.Markdown( "The entity visualizer highlights named entities and their labels in a text.") ent_input = gr.CheckboxGroup( DEFAULT_ENTS, value=DEFAULT_ENTS) ent_output = gr.HTML(value=entity( DEFAULT_TEXT, DEFAULT_ENTS, DEFAULT_MODEL)) ent_button = gr.Button("Generate Entity Recognizer", variant="primary") # with gr.Row(): # with gr.Column(): # ent_button = gr.Button("Generate Entity Recognizer", variant="primary") # with gr.Column(): # ent_download_button = gr.HTML(value=download_svg(ent_output.value)) # with gr.Column(): # gr.Markdown(" ") # with gr.Column(): # gr.Markdown(" ") with gr.Box(): with gr.Column(): gr.Markdown( "## [Token Properties](https://spacy.io/usage/linguistic-features)") gr.Markdown( "When you put in raw text to spaCy, it returns a Doc object with different linguistic features") with gr.Row(): with gr.Column(): tok_input = gr.CheckboxGroup( DEFAULT_TOK_ATTR, value=DEFAULT_TOK_ATTR) with gr.Column(): gr.Markdown("") tok_output = gr.Dataframe(headers=DEFAULT_TOK_ATTR, value=default_token( DEFAULT_TEXT, DEFAULT_TOK_ATTR, DEFAULT_MODEL), overflow_row_behaviour="paginate") tok_button = gr.Button("Generate Token Properties", variant="primary") with gr.Box(): with gr.Column(): gr.Markdown( "## [Word and Phrase Similarity](https://spacy.io/usage/linguistic-features#vectors-similarity)") gr.Markdown( "Words and spans have similarity ratings based off of their word vectors, or word embeddings") with gr.Row(): with gr.Column(): sim_text1 = gr.Textbox( value="Apple", label="Word 1", interactive=True,) with gr.Column(): sim_text2 = gr.Textbox( value="U.K. startup", label="Word 2", interactive=True,) with gr.Column(): sim_output = gr.Textbox( label="Similarity Score", value="0.12") with gr.Column(): gr.Markdown("") sim_random_button = gr.Button("Generate random words") sim_button = gr.Button("Generate similarity", variant="primary") with gr.Box(): with gr.Column(): gr.Markdown( "## [Spans](https://spacy.io/usage/visualizers#span)") gr.Markdown( "The span visualizer highlights overlapping spans in a text.") with gr.Row(): with gr.Column(): span1 = gr.Textbox( label="Span 1", value="U.K. startup", placeholder="Input a part of the sentence") with gr.Column(): label1 = gr.Textbox(value="ORG", label="Label for Span 1") with gr.Column(): gr.Markdown("") with gr.Column(): gr.Markdown("") with gr.Row(): with gr.Column(): span2 = gr.Textbox( label="Span 2", value="U.K.", placeholder="Input another part of the sentence") with gr.Column(): label2 = gr.Textbox(value="GPE", label="Label for Span 2") with gr.Column(): gr.Markdown("") with gr.Column(): gr.Markdown("") span_output = gr.HTML(value=span( DEFAULT_TEXT, "U.K. startup", "U.K.", "ORG", "GPE", DEFAULT_MODEL)) span_button = gr.Button("Generate Spans", variant="primary") # with gr.Row(): # with gr.Column(): # span_button = gr.Button("Generate Spans", variant="primary") # with gr.Column(): # span_download_button = gr.HTML(value=download_svg(span_output.value)) # with gr.Column(): # gr.Markdown(" ") # with gr.Column(): # gr.Markdown(" ") model_input.change(get_text, inputs=[model_input], outputs=text_input) button.click(dependency, inputs=[ text_input, col_punct, col_phrase, compact, bg, text, model_input], outputs=[depen_output, dep_download_button]) button.click( entity, inputs=[text_input, ent_input, model_input], outputs=[ent_output]) button.click( token, inputs=[text_input, tok_input, model_input], outputs=tok_output) button.click(vectors, inputs=[sim_text1, sim_text2, model_input], outputs=sim_output) button.click( span, inputs=[text_input, span1, span2, label1, label2, model_input], outputs=[span_output]) dep_button.click(dependency, inputs=[ text_input, col_punct, col_phrase, compact, bg, text, model_input], outputs=[depen_output, dep_download_button]) ent_button.click( entity, inputs=[text_input, ent_input, model_input], outputs=[ent_output]) tok_button.click( token, inputs=[text_input, tok_input, model_input], outputs=[tok_output]) sim_button.click(vectors, inputs=[ sim_text1, sim_text2, model_input], outputs=sim_output) span_button.click( span, inputs=[text_input, span1, span2, label1, label2, model_input], outputs=[span_output]) sim_random_button.click(random_vectors, inputs=[text_input, model_input], outputs=[ sim_output, sim_text1, sim_text2]) demo.launch()