app.py

from difflib import Differ

import gradio as gr
from transformers import pipeline

pipe = pipeline("summarization", "dominguesm/positive-reframing-en")


def predict(text, operation):

    try:
        res = pipe(f"[{operation}]: {text}", max_length=124)
    except Exception as e:
        return e

    d = Differ()
    return (
        res[0]["summary_text"],
        [
            (token[2:], token[0] if token[0] != " " else None)
            for token in d.compare(text, res[0]["summary_text"])
        ],
    )
    # return res[0]["summary_text"]


iface = gr.Interface(
    title="Positive Reframing EN",
    description="This model is a T5 adjusted to the sentiment transfer task, where the objective is to reverse the sentiment polarity of a text without contradicting the original meaning. Positive reframing induces a complementary positive viewpoint (e.g. glass-half-full) escaping negative patterns. More info [here](https://huggingface.co/dominguesm/positive-reframing-en).",
    fn=predict,
    inputs=[
        gr.Textbox(
            lines=1,
            placeholder=(
                f"Pensar no meu futuro me faz querer viver numa ilha sozinha para sempre"
            ),
        ),
        gr.Radio(
            [
                "growth",
                "impermanence",
                "neutralizing",
                "optimism",
                "self_affirmation",
                "thankfulness",
            ]
        ),
    ],
    outputs=[
        gr.Textbox(label="Generated Text"),
        gr.HighlightedText(
            label="Diff",
            combine_adjacent=True,
        ).style(color_map={"+": "green", "-": "red"}),
    ],
    examples=[
        [
            "You know I really don't care about the power struggle between the papacy and secular authority in the medieval ages. stupid",
            "growth",
        ],
        [
            "thinking about my future makes me want to go live on a island alone forever. annoyed",
            "optimism",
        ],
        [
            "Who would have ever guessed that it would be so freaking hard to get three different grades from two different schools together.",
            "thankfulness",
        ],
    ],
)

iface.launch()








import streamlit as st

# Dataset
examples = [
    [
        "The power struggle between the papacy and secular authority in the medieval ages is as unimportant to me as a drop of water in the ocean. (simile)",
        "growth",
    ],
    [
        "Contemplating my future feels like being engulfed by the urge to escape to a secluded island, forever. (metaphor)",
        "optimism",
    ],
    [
        "Who would have thought that uniting three different grades from two different schools would be as tough as nailing jelly to a wall? (simile)",
        "thankfulness",
    ],
    [
        "Her laughter was like the tinkling of silver bells, filling the room with joy. (simile)",
        "happiness",
    ],
    [
        "The thunder roared and boomed, striking fear in the hearts of those who heard it. (onomatopoeia)",
        "courage",
    ],
]

language_features = [
    "Metaphor",
    "Simile",
    "Onomatopoeia",
    "Alliteration",
    "Assonance",
    "Hyperbole",
    "Personification",
    "Oxymoron",
    "Paradox",
    "Pun",
    "Irony",
    "Sarcasm",
    "Allusion",
    "Imagery",
    "Symbolism",
    "Anaphora",
    "Epistrophe",
    "Parallelism",
    "Euphemism",
    "Synecdoche",
]

# Streamlit app
st.title("Language Feature Emoji Reference")

# Table of buttons
table_data = [language_features[i:i + 3] for i in range(0, len(language_features), 3)]

for row in table_data:
    row_buttons = st.beta_columns(len(row))
    for i, feature in enumerate(row):
        if row_buttons[i].button(feature):
            for example in examples:
                if feature.lower() in example[0]:
                    st.write(f"**{feature}:** {example[0]}")
                    st.write(f"**Emoji:** {example[1]}")
                    st.write("---")