Cloudy1225/stackoverflow-sentiment-analysis

1import csv2import gradio as gr3import pandas as pd4from sentiment_analyser import RandomAnalyser, RoBERTaAnalyser, ChatGPTAnalyser5import matplotlib.pyplot as plt6from sklearn.metrics import ConfusionMatrixDisplay, confusion_matrix7 8 9def plot_bar(value_counts):10    fig, ax = plt.subplots(figsize=(6, 6))11    value_counts.plot.barh(ax=ax)12    ax.bar_label(ax.containers[0])13    plt.title('Frequency of Predictions')14    return fig15 16 17def plot_confusion_matrix(y_pred, y_true):18    cm = confusion_matrix(y_true, y_pred, normalize='true')19    fig, ax = plt.subplots(figsize=(6, 6))20    labels = []21    for label in SENTI_MAPPING.keys():22        if (label in y_pred.values) or (label in y_true.values):23            labels.append(label)24    disp = ConfusionMatrixDisplay(confusion_matrix=cm,25                                  display_labels=labels)26    disp.plot(cmap="Blues", values_format=".2f", ax=ax, colorbar=False)27    plt.title("Normalized Confusion Matrix")28    return fig29 30 31def classify(num: int):32    samples_df = df.sample(num)33    X = samples_df['Text'].tolist()34    y = samples_df['Label']35    roberta = MODEL_MAPPING[OUR_MODEL]36    y_pred = pd.Series(roberta.predict(X), index=samples_df.index)37    samples_df['Predict'] = y_pred38    bar = plot_bar(y_pred.value_counts())39    cm = plot_confusion_matrix(y_pred, y)40    plt.close()41    return samples_df, bar, cm42 43 44def analysis(Text):45    keys = []46    values = []47    for name, model in MODEL_MAPPING.items():48        keys.append(name)49        values.append(SENTI_MAPPING[model.predict([Text])[0]])50    return pd.DataFrame([values], columns=keys)51 52 53def analyse_file(file):54    output_name = 'output.csv'55    with open(output_name, mode='w', newline='') as output:56        writer = csv.writer(output)57        header = ['Text', 'Label']58        writer.writerow(header)59        model = MODEL_MAPPING[OUR_MODEL]60        with open(file.name) as f:61            for line in f:62                text = line[:-1]63                sentiment = model.predict([text])64                writer.writerow([text, sentiment[0]])65    return output_name66 67 68MODEL_MAPPING = {69    'Random': RandomAnalyser(),70    'RoBERTa': RoBERTaAnalyser(),71    'ChatGPT': RandomAnalyser(),72}73 74OUR_MODEL = 'RoBERTa'75 76SENTI_MAPPING = {77    'negative': '😭',78    'neutral': '😶',79    'positive': '🥰'80}81 82TITLE = "Sentiment Analysis on Software Engineer Texts"83 84DESCRIPTION = {85    'en': (86        "This is the demo page for our model: "87        "[Cloudy1225/stackoverflow-roberta-base-sentiment]"88        "(https://huggingface.co/Cloudy1225/stackoverflow-roberta-base-sentiment)."89    ),90    'zh': (91        "这里是第16组“睿王和他的五个小跟班”软工三迭代三模型演示页面。"92        "模型链接：[Cloudy1225/stackoverflow-roberta-base-sentiment]"93        "(https://huggingface.co/Cloudy1225/stackoverflow-roberta-base-sentiment)."94    )95}96 97PROMPT1 = {98    'en': (99        "Enter text in the left text box and press Enter, and the sentiment analysis results will be output on the right. "100        "Here, we present three types of results, which come from random, our model, and ChatGPT."101    ),102    'zh': (103        "在左侧文本框中输入文本并按回车键，右侧将输出情感分析结果。"104        "这里我们展示了三种结果，分别是随机结果、模型结果和 ChatGPT 结果。"105    )106}107 108PROMPT2 = {109    'en': (110        "Upload a txt/csv file in the left file box, and the model will perform sentiment analysis on each line of the input text. "111        "You can download the output file on the right. "112        "The output file will be in CSV format with two columns: the original text, and the classification results."113    ),114    'zh': (115        "在左侧文件框中上传 txt/csv 文件，模型会对输入文本的每一行当作一个文本进行情感分析。"116        "可以在右侧下载输出文件，输出文件为两列 csv 格式，第一列为原始文本，第二列为分类结果。"117    )118}119 120PROMPT3 = {121    'en': (122        "Here we evaluate our model on the StackOverflow4423 dataset. "123        "Sliding the slider will sample a specified number of samples from the StackOverflow4423 dataset and predict their sentiment labels. "124        "Based on the prediction results, a label distribution chart and a confusion matrix will be plotted."125    ),126    'zh': (127        "这里是在 StackOverflow4423 数据集上评估我们的模型。"128        "滑动 Slider，将会从 StackOverflow4423 数据集中抽样出指定数量的样本，预测其情感标签。"129        "并根据预测结果绘制标签分布图和混淆矩阵。"130    )131}132 133DEFAULT_LANG = 'en'134 135MAX_SAMPLES = 64136 137df = pd.read_csv('./SOF4423.csv')138 139 140def set_language(lang):141    return DESCRIPTION[lang], PROMPT1[lang], PROMPT2[lang], PROMPT3[lang]142 143 144with gr.Blocks(title=TITLE) as demo:145    with gr.Row():146        with gr.Column():147            gr.HTML(f"<H1>{TITLE}</H1>")148        with gr.Column(min_width=160):149            language_selector = gr.Radio(150                ['en', 'zh'], label="Select Language", value=DEFAULT_LANG,151                interactive=True, show_label=False, container=False152            )153 154    description = gr.Markdown(DESCRIPTION[DEFAULT_LANG])155    gr.HTML("<H2>Model Inference</H2>")156    prompt1 = gr.Markdown(PROMPT1[DEFAULT_LANG])157    with gr.Row():158        with gr.Column():159            text_input = gr.Textbox(label='Input',160                                    placeholder="Enter a positive or negative sentence here...")161        with gr.Column():162            senti_output = gr.Dataframe(type="pandas", value=[['😋', '😋', '😋']],163                                        headers=list(MODEL_MAPPING.keys()), interactive=False)164    text_input.submit(analysis, inputs=text_input, outputs=senti_output, show_progress='full')165 166    prompt2 = gr.Markdown(PROMPT2[DEFAULT_LANG])167    with gr.Row():168        with gr.Column():169            file_input = gr.File(label='File',170                                 file_types=['.txt', '.csv'])171        with gr.Column():172            file_output = gr.File(label='Output')173    file_input.upload(analyse_file, inputs=file_input, outputs=file_output)174 175    gr.HTML("<H2>Model Evaluation</H2>")176    prompt3 = gr.Markdown(PROMPT3[DEFAULT_LANG])177    input_models = list(MODEL_MAPPING)178    input_n_samples = gr.Slider(179        minimum=4,180        maximum=MAX_SAMPLES,181        value=8,182        step=4,183        label='Number of samples'184    )185 186    with gr.Row():187        with gr.Column():188            bar_plot = gr.Plot(label='Predictions Frequency')189        with gr.Column():190            cm_plot = gr.Plot(label='Confusion Matrix')191 192    with gr.Row():193        dataframe = gr.Dataframe(type="pandas", wrap=True, headers=['Text', 'Label', 'Predict'])194 195    input_n_samples.change(fn=classify, inputs=input_n_samples, outputs=[dataframe, bar_plot, cm_plot])196 197    language_selector.change(fn=set_language, inputs=language_selector,198                             outputs=[description, prompt1, prompt2, prompt3])199 200    demo.launch()201