EmoMosaic-large

EmoMosaic-large is a well-competent model designed for classifying emotions in text, demonstrating strong performance across multiple domains. In fact, it outperforms recent state-of-the-art models on SemEval-2018 Task 1: Affect in Tweets and GoEmotions datasets. The model was developed as a part of my master's thesis.

Author

Author: Bc. Vít Tlustoš ([email protected])

Supervisor: doc. Malik Aamir Saeed Ph.D

Thesis Text

Once the thesis has been defended, the text will be accessible at https://www.vut.cz/studenti/zav-prace/detail/153407. You are welcome to read it. Should you have any questions, please don't hesitate to contact me via the provided email.

Demo Application

As a part of the solution, we developed a Gradio application and deployed it on the Hugging Face Spaces platform. Once the thesis is made public, you can access it at: https://huggingface.co/spaces/vtlustos/EmoMosaic-space. This allows anyone to experiment with the models easily without requiring any technical skills or setup.

Models

To utilize these models within your application, first install all the necessary dependencies.

pip install torch
pip install transformers
pip install datasets

To utilize these models within your application, integrate following code and format your samples as context</s><s>sentence. The context is optional and represents sentences preceding the sentence to be classified, while sentence refers to the actual sentence undergoing classification. This example demonstrates how to use the EmoMosaic-base model. If you prefer to use its larger counterpart, replace vtlustos/EmoMosaic-base with vtlustos/EmoMosaic-large.

import torch
from transformers import RobertaTokenizer
from transformers import RobertaForSequenceClassification

# 1. initialize the model
tokenizer = RobertaTokenizer.from_pretrained(
    "vtlustos/EmoMosaic-base"
)
model = RobertaForSequenceClassification.from_pretrained(
    "vtlustos/EmoMosaic-base"
).to('cuda:0')

# 2. tokenize the sentences
tokens = tokenizer(
    [
        "All your work was lost when the computer crashed.</s><s>Oh my god. I spent a whole week on that."
    ],
    truncation=True,
    padding=True,
    return_tensors = "pt"
)    

# 3. make the prediction
with torch.no_grad():
    logits = model(
        tokens["input_ids"].to('cuda:0'), 
        tokens["attention_mask"].to('cuda:0')
    ).logits

# 4. convert to probabilities
preds = torch.sigmoid(logits)

print(preds)

After executing the code, you will receive a tensor with dimensions [S,E], where S represents the number of samples and E denotes the number of emotions. To associate individual probabilities with their respective emotions, use to the dictionary provided below:

ix2label = {
    "0": "admiration",
    "1": "amusement",
    "2": "anger",
    "3": "annoyance",
    "4": "anticipation",
    "5": "approval",
    "6": "caring",
    "7": "confusion",
    "8": "curiosity",
    "9": "desire",
    "10": "disappointment",
    "11": "disapproval",
    "12": "disgust",
    "13": "embarrassment",
    "14": "excitement",
    "15": "fear",
    "16": "gratitude",
    "17": "grief",
    "18": "happiness",
    "19": "joy",
    "20": "love",
    "21": "nervousness",
    "22": "optimism",
    "23": "pessimism",
    "24": "pride",
    "25": "realization",
    "26": "relief",
    "27": "remorse",
    "28": "sadness",
    "29": "surprise",
    "30": "trust"
}

Results

Here we present a brief overview of the results. For an in-depth analysis and discussion, please refer to the text of the thesis. The analysis covers model training, comparisons with other methods, assessments of performance at the level of individual categories, calibration, and qualitative evaluations across various scenarios.

SemEval-2018 Task 1: Affect in Tweets

Model	Accuracy	P (macro)	R (macro)	F1 (macro)	P (micro)	R (micro)	F1 (micro)
EmoMosaic-base	20.65	54.96	62.58	58.44	64.63	73.62	68.83
EmoMosaic-large	22.49	57.97	64.12	60.72	67.44	75.27	71.14

Note: P and R denote precision and recall, respectively. Results are shown for our top-performing models measured on the test set of the SemEval-2018 Task 1: Affect in Tweets dataset.

GoEmotions

Model	Accuracy	P (macro)	R (macro)	F1 (macro)	P (micro)	R (micro)	F1 (micro)
EmoMosaic-base	46.47	51.41	57.81	53.72	52.70	62.53	57.19
EmoMosaic-large	46.67	51.35	58.34	53.93	52.86	63.39	57.65

Note: P and R denote precision and recall, respectively. Results are shown for our two top-performing models measured on the test set of the GoEmotions dataset.

XED

Model	Accuracy	P (macro)	R (macro)	F1 (macro)	P (micro)	R (micro)	F1 (micro)
EmoMosaic-base	51.78	48.47	63.00	54.67	48.62	63.86	55.21
EmoMosaic-large	52.59	50.35	66.54	57.19	50.43	67.43	57.70

Note: P and R denote precision and recall, respectively. Results are shown for our two top-performing models measured on the test set of the XED dataset.

DailyDialog

Model	Accuracy	P (macro)	R (macro)	F1 (macro)	P (micro)	R (micro)	F1 (micro)
EmoMosaic-base	84.85	46.34	49.60	46.94	53.44	64.81	58.57
EmoMosaic-large	85.05	47.20	53.80	49.65	54.24	68.77	60.65

Note: P and R denote precision and recall, respectively. Results are shown for our two top-performing models measured on the test set of the DailyDialog dataset.

Per-Emotion Performance

EmoMosaic-base

Emotion	Precision	Recall	F1
admiration	63.82	80.16	71.06
amusement	74.11	94.32	83.00
anger	63.46	74.08	68.36
annoyance	35.15	44.37	39.23
anticipation	39.09	55.15	45.75
approval	43.40	45.87	44.60
caring	45.67	42.96	44.27
confusion	36.10	56.86	44.16
curiosity	48.48	67.25	56.34
desire	53.09	51.81	52.44
disappointment	35.57	35.10	35.33
disapproval	40.00	49.44	44.22
disgust	62.05	71.31	66.36
embarrassment	57.69	40.54	47.62
excitement	37.40	44.66	40.71
fear	61.93	68.69	65.13
gratitude	93.29	90.91	92.09
grief	66.67	66.67	66.67
happiness	58.10	70.76	63.81
joy	73.43	81.18	77.11
love	64.95	73.74	69.07
nervousness	33.33	43.48	37.74
optimism	64.33	76.00	69.68
pessimism	42.31	52.80	46.98
pride	66.67	37.50	48.00
realization	32.71	24.14	27.78
relief	55.56	45.45	50.00
remorse	55.56	89.29	68.49
sadness	58.65	70.14	63.88
surprise	40.02	51.29	44.96
trust	35.33	47.01	40.34

EmoMosaic-large

Emotion	Precision	Recall	F1
admiration	65.25	79.37	71.62
amusement	73.87	93.18	82.41
anger	64.29	76.00	69.66
annoyance	33.81	44.06	38.26
anticipation	42.10	57.99	48.78
approval	42.66	44.73	43.67
caring	40.26	45.93	42.91
confusion	38.76	52.94	44.75
curiosity	48.40	74.65	58.73
desire	65.08	49.40	56.16
disappointment	34.36	37.09	35.67
disapproval	39.14	47.94	43.10
disgust	63.62	72.30	67.68
embarrassment	58.33	37.84	45.90
excitement	39.82	43.69	41.67
fear	64.22	71.24	67.55
gratitude	91.01	92.05	91.53
grief	66.67	66.67	66.67
happiness	58.21	75.23	65.63
joy	74.55	83.53	78.78
love	64.13	76.13	69.62
nervousness	42.86	39.13	40.91
optimism	66.98	79.38	72.66
pessimism	43.66	47.73	45.61
pride	63.64	43.75	51.85
realization	34.29	24.83	28.80
relief	33.33	36.36	34.78
remorse	57.78	92.86	71.23
sadness	61.08	72.67	66.37
surprise	44.02	55.67	49.16
trust	40.59	48.26	44.09