app.py

# This is our main interface library
# For main things
import types
import bibtexparser
import ethnicolr
import genderComputer
import gender_guesser.detector
import nameparser
import numpy
import pandas
import plotly.express
import streamlit
import st_aggrid


class References(object):
    def __init__(self, reference_text):
        self.gender_options = ['male', 'mostly_male', 'andy', 'mostly_female', "female", "unknown",
                               "first_name_initial"]
        self.gender_results = {key: 0 for key in self.gender_options}
        self.race_options = ['pctwhite', 'pctblack', 'pctapi', 'pctaian', 'pct2prace', 'pcthispanic', 'race_unknown']
        self.ethnicity_results = {key: 0 for key in self.race_options}
        self.raw_results = pandas.DataFrame(columns=["First Name", "Last Name", "Title"])

        # Parse names from input
        self.reference_text = reference_text
        self.references = bibtexparser.loads(reference_text)
        for paper in self.references.entries:
            if "author" in paper:
                authors = paper["author"].split(' and ')
                for person in authors:
                    name = nameparser.HumanName(person)
                    self.raw_results.loc[len(self.raw_results.index)] = [name.first, name.last, paper['title']]

    def infer_ethnicity(self):
        if ethnicity_model == "ethnicolr - census data":
            other = ethnicolr.pred_census_ln(self.raw_results, 'Last Name', 2010)
            self.raw_results['Most Likely Ethnicity'] = other['race']

        elif ethnicity_model == "ethnicolr - wikipedia data":
            other = ethnicolr.pred_wiki_name(self.raw_results, 'Last Name', 'First Name')
            self.raw_results['Most Likely Ethnicity'] = other['race']
            self.raw_results.drop(columns=['__name'])

        elif ethnicity_model == "ethnicolr - Florida registration data":
            other = ethnicolr.pred_fl_reg_name_five_cat(self.raw_results, 'Last Name', 'First Name')
            self.raw_results['Most Likely Ethnicity'] = other['race']
            self.raw_results.drop(columns=['__name'])

        elif ethnicity_model == "ethnicolr - North Carolina data":
            other = ethnicolr.pred_nc_reg_name(self.raw_results, 'Last Name', 'First Name')
            self.raw_results['Most Likely Ethnicity'] = other['race']
            self.raw_results.drop(columns=['__name'])

        for i in self.raw_results['Most Likely Ethnicity']:
            self.ethnicity_results[i] = self.ethnicity_results.get(i, 0) + 1

    def infer_gender(self):

        most_likely_gender = []
        if gender_model == "gender_guesser":
            d = gender_guesser.detector.Detector()
            for name in self.raw_results['First Name']:
                if (len(name) == 2 and name[1] == '.') or len(name) == 1:
                    most_likely_gender.append("first_name_initial")
                else:
                    most_likely_gender.append(d.get_gender(name))
        elif gender_model == "genderComputer":
            gc = genderComputer.GenderComputer()
            for idx in self.raw_results.index:
                fn = self.raw_results['First Name'][idx]
                if (len(fn) == 2 and fn[1] == '.') or len(fn) == 1:
                    most_likely_gender.append("first_name_initial")
                else:
                    most_likely_gender.append(
                        gc.resolveGender(self.raw_results['First Name'][idx] + " " + self.raw_results['Last Name'][idx],
                                         None))
                    if most_likely_gender[-1] is None:
                        most_likely_gender[-1] = "unknown"
        self.raw_results['Most Likely Gender'] = most_likely_gender
        for i in most_likely_gender:
            self.gender_results[i] = self.gender_results.get(i, 0) + 1


label_to_gender = {'male': "Very Likely Male",
                   'mostly_male': "Likely Male",
                   'andy': "Hard to Tell",
                   'unisex': "Hard to Tell",
                   'mostly_female': "Likely Female",
                   "female": "Very Likely Female",
                   "unknown": "Unknown (model inconclusive)",
                   "first_name_initial": "Unknown (first name initial only)"}

label_to_ethnicity = {
    'white': 'White',
    'black': 'Black',
    'api': 'Asian or Pacific Islander',
    'hispanic': 'Hispanic',
    'pctwhite': 'White',
    'pctblack': 'Black',
    'pctapi': 'Asian or Pacific Islander',
    'pctaian': 'American Indian or Alaskan Native',
    'pct2prace': 'Two or more races',
    'pcthispanic': 'Hispanic',
    'race_unknown': 'Unknown (not found in database)'}

ethnicity_to_label = {v: k for k, v in label_to_ethnicity.items()}
gender_to_label = {v: k for k, v in label_to_gender.items()}

def make_table():
    if 'table_data' in streamlit.session_state:
        df = streamlit.session_state['table_data']
    else:
        refs = References(streamlit.session_state.bib)
        refs.infer_gender()
        refs.infer_ethnicity()

        df = refs.raw_results
        df = df.replace({"Most Likely Ethnicity": label_to_ethnicity})
        df = df.replace({"Most Likely Gender": label_to_gender})
        df = df.sort_values(["Last Name", "First Name"])
        df = df.reset_index(drop=True)

    gb = st_aggrid.GridOptionsBuilder.from_dataframe(df)
    gb.configure_default_column(editable=True)

    gb.configure_column('Most Likely Ethnicity',
                        cellEditor='agRichSelectCellEditor',
                        cellEditorParams={'values': list(label_to_ethnicity.values())},
                        cellEditorPopup=True
                        )

    gb.configure_column('Most Likely Gender',
                        cellEditor='agRichSelectCellEditor',
                        cellEditorParams={'values': list(label_to_gender.values())},
                        cellEditorPopup=True
                        )

    gb.configure_column('Title',
                        editable=False
                        )

    response = st_aggrid.AgGrid(
        data=df,
        gridOptions=gb.build(),
        fit_columns_on_grid_load=True,
        update_mode=st_aggrid.GridUpdateMode.VALUE_CHANGED,
        height=400,
    )

    streamlit.session_state['table_data'] = response.data
    # if response.column_state:
    #     streamlit.experimental_rerun()


# Define a function for addition
def make_results():
    data = streamlit.session_state['table_data']
    refs = types.SimpleNamespace(
        ethnicity_results=data['Most Likely Ethnicity'].value_counts().to_dict(),
        gender_results=data['Most Likely Gender'].value_counts().to_dict(),
    )

    plt1 = plotly.express.pie(
        names=list(refs.ethnicity_results.keys()),
        values=refs.ethnicity_results.values(),
        hole=0.5,
    )
    plt2 = plotly.express.pie(
        names=list(refs.gender_results.keys()),
        values=refs.gender_results.values(),
        hole=0.5,
    )
    plt1.update_layout(legend=dict(orientation="h"))
    plt2.update_layout(legend=dict(orientation="h"))

    col1, col2, col3, col4 = streamlit.columns(4)
    col1.metric("Ethnicity Unknown",
                "{pct:.0%}".format(pct=len(data[data['Most Likely Ethnicity'].str.contains('known')]) /
                                       len(data))
                )

    col2.metric("Gender Unknown",
                "{pct:.0%}".format(pct=len(data[data['Most Likely Gender'].str.contains('known|Hard')]) /
                                       len(data))
                )

    col3.metric("G or E Unknown",
                "{pct:.0%}".format(pct=len(data[numpy.logical_or(data['Most Likely Gender'].str.contains('known|Hard'),
                                                                 data['Most Likely Ethnicity'].str.contains(
                                                                     'known'))]) /
                                       len(data))
                )

    col4.metric("G and E Unknown",
                "{pct:.0%}".format(pct=len(data[numpy.logical_and(data['Most Likely Gender'].str.contains('known|Hard'),
                                                                  data['Most Likely Ethnicity'].str.contains(
                                                                      'known'))]) /
                                       len(data))
                )

    streamlit.plotly_chart(plt1, use_container_width=True)
    streamlit.plotly_chart(plt2, use_container_width=True)


streamlit.title("Welcome and thanks for coming")
streamlit.markdown("""Simply put, many people often cite people that are like them. This is a problem because academia has historically been white male dominated, leading to the suppression of marginalized voices. If your citations are biased towards people who look like you, then you are missing out on high-quality work.
    
Its important to note that using this site is not a replacement for truly being diligent and engaged in citing diverse voices. Rather, this site is just a place to start, and hopefully the first step in your journey of citing more diversely. To learn more about your duty to dismantle institutional oppression through your citation practices, read up here:
    
- [Cite Black Women](https://www.citeblackwomencollective.org)
- [The Racial Politics of Citation](https://www.insidehighered.com/advice/2018/04/27/racial-exclusions-scholarly-citations-opinion")
- [Inclusive Citation: How Diverse Are Your References?](https://blog.mahabali.me/writing/inclusive-citation-how-diverse-are-your-references/")
  
""")

streamlit.markdown("To use our tool, copy and paste your references in the box below and click on the "
                   "`Analyze` button.")

filler = """@article{Raina2019,
    author = {Raina, Ayush and McComb, Christopher and Cagan, Jonathan},
    title = {Learning to Design From Humans: Imitating Human Designers Through Deep Learning},
    journal = {Journal of Mechanical Design},
    volume = {141},
    number = {11},
    year = {2019},
    month = {09},
    issn = {1050-0472},
    doi = {10.1115/1.4044256}
}

@article{Williams2019,
    author = {Williams, Glen and Meisel, Nicholas A. and Simpson, Timothy W. and McComb, Christopher},
    title = {Design Repository Effectiveness for 3D Convolutional Neural Networks: Application to Additive Manufacturing},
    journal = {Journal of Mechanical Design},
    volume = {141},
    number = {11},
    year = {2019},
    month = {09},
    issn = {1050-0472},
    doi = {10.1115/1.4044199}
}"""
if "bib" in streamlit.session_state:
    filler = streamlit.session_state["bib"]

streamlit.text_area(".bibtex only for now, sorry!", filler, key="bib", height=250)
details = streamlit.sidebar
details.markdown("## Settings")
details.markdown("The following dropdowns allow you to select the specific models that are used for analysis.")
details.markdown("When it comes to gender inference, you have two models to choose from: [`gender_guesser`](https://github.com/lead-ratings/gender-guesser) (which is primarily trained on European names), and [`genderComputer`](https://github.com/tue-mdse/genderComputer) (which draws on more global data).")
gender_model = details.selectbox("Gender Inference Model", ("gender_guesser", "genderComputer"))
details.markdown("The ethnicity inference models are all drawn from the [`ethnicolr`](https://github.com/appeler/ethnicolr) library.  The models are differentiated by the data they are trained on.")
ethnicity_model = details.selectbox("Ethnicity Inference Model", ("ethnicolr - census data",
                                                                  "ethnicolr - wikipedia data",
                                                                  "ethnicolr - North Carolina data",
                                                                  "ethnicolr - Florida registration data"))

placeholder = streamlit.empty()
time_to_analyze = placeholder.button("Analyze")
if time_to_analyze or 'already_analyzed' in streamlit.session_state:
    streamlit.session_state['already_analyzed'] = True
    placeholder.empty()
    with streamlit.spinner("Analyzing..."):
        make_table()
        streamlit.markdown("These tabs summarize your results with a variety of visualizations and statistics.")
        make_results()