Multilingual Sentiment Analysis

Prof K. Anusha
Ashley Edgar Dcunha, Harshith Chemela, Rachapally Pavan Kumar, Sasidhar Kashyap, Vellore Anand Kumar
Sahil

ABSTRACT
People's sentiments are known to have a large impact on changes in stock prices, products sales, and trends. On the
internet, people share their thoughts in many different languages. This is why it's important to create a way to
understand how people feel in different languages when analyzing web text. Most of the text also include emojis. So,
it's really important to take emojis into account when figuring our how people feel online. In our project, we're working
on a method that can understand how people feel in many languages and even consider the emotions shown by emojis.
This way, we can better understand what people are saying online, no matter what language they're use. This will help
us make better predictions and decisions in fields of sentiment analysis.

I. INTRODUCTION properly identifying emotions in text.

Our project focuses on understanding how people express
their feelings in different languages, including the There are many more uses such as Customer Experience
and Marketing Research, Psychological and mental
emotions conveyed by emojis. Emojis are those small health analysis, Human-Computer interaction, Social
pictures you often see in text messages or social media media analysis and monitoring, Content generalization
posts that show feelings like happiness, sadness, or and personalization and Educational Applications.
excitement.
As our world becomes more connected through the Amira M. Idrees, et al.'s "Emotion Detection in Text: A
internet, it's important to be able to understand what Review" was published in 2017.This review paper offers
a thorough examination of machine learning-based
people are saying and feeling, especially in different
emotion detection methods in text. It addresses many
languages. Traditional ways of analyzing how people feel strategies, including deep learning-based, supervised, and
usually look at the words they use, but emojis add another unsupervised techniques. Various features, datasets,
layer of meaning that we need to consider, especially assessment metrics, and obstacles in the field of emotion
when people are using different languages. detection are also covered in the study.
Our paper looks at how we can build computer programs
that can understand both the words people use and the S. Aarthi published "A Review on Emotion Detection
Techniques using Textual Data" in 2020. This review
emojis they include, even when they're writing in
study provides an in-depth analysis of textual data-based
different languages. This kind of understanding can help emotion detection methods. It discusses deep learning
us get a better grasp of how people feel and what they're methods including CNN, RNN, and Transformer models
talking about, no matter what language they're using. in addition to more conventional machine learning
strategies like SVM, Naive Bayes, and decision trees. The
II. SURVEY study also looks at several feature extraction techniques
and field-relevant datasets.
Sentiment analysis, which involves locating and
K. Kalimuthu's "Emotion Detection from Text: A
classifying the sentiment or opinion represented in text, Survey" was published in 2019.This survey paper offers
heavily relies on emotion detection. Sentiment analysis
a thorough analysis of machine learning-based emotion
systems can offer insights into consumer feedback, social recognition from text. It covers several topics, including
media posts, product evaluations, and public opinion by
feature extraction techniques, classification algorithms,
and assessment measures. The study also covers the Language Understanding: We use natural language
field's difficulties, uses, and future directions. processing techniques to teach the computer how to
understand the different languages in our dataset,
Emotion detection of text using machine learning has
including their grammar, vocabulary, and how emojis are
been a widely studied topic, and numerous research
papers have been published in this area. Some of the used.
challenges and future directions are as follows
Dealing with sarcasm, irony, and implicit emotions, Emoji Analysis: We develop a way for the computer to
handling noisy or imbalanced datasets, and handling recognize and interpret emojis in the text, considering
cultural and contextual variances are difficulties in their meanings in different cultures and contexts.
emotion identification of text using machine learning.
Future research initiatives should focus on the
Sentiment Analysis: Using the processed text and emoji
interpretability and explain ability of models, multi-
modal techniques that mix text with additional modalities data, we train the computer to recognize the emotions
(such as visuals and audio), and real-time emotion expressed in the text, taking into account both the words
recognition in dynamic text streams. used and the emojis included.

III. CHALLENGES V. KEY FEATURES

The various challenges encountered in the existing Multilingual Analysis: Our project focuses on analyzing
models include sentiments expressed in multiple languages, allowing for
Data Collection: Acquiring a diverse and representative a broader understanding of emotions across different
dataset of multilingual text containing emojis can be linguistic communities.
challenging, as it requires access to a wide range of Emoji Integration: We incorporate the analysis of emojis
languages and cultural contexts. into our sentiment analysis model, recognizing the
Language Complexity: Different languages have unique significance of these symbols in modern digital
grammatical structures, idiomatic expressions, and communication.
linguistic nuances, making it challenging to develop a Natural Language Processing (NLP) Techniques: We
model that can effectively analyze sentiment across utilize advanced NLP techniques to effectively interpret
diverse linguistic backgrounds. both textual data and emojis, enhancing the accuracy of
Emoji Interpretation: Emojis can be highly context- our sentiment analysis.
dependent and culturally specific, requiring a deep Practical Applications: The insights gained from our
understanding of their meanings and usage in different project can be applied to various real-world scenarios,
languages and cultures. such as understanding customer feedback, analyzing
Multilingual Training: Training a model to accurately social media trends, and gauging public opinion across
analyze sentiments in multiple languages requires a large different language communities.
amount of labeled data for each language, which may not
always be readily available.

IV. PROPOSED METHOD VI. CONCLUSION

In our project, we've shown that it's possible to build
Data Collection: We gather a wide range of text samples sentiment analysis model that can understand how people
from different languages that contain emojis, ensuring we feel in different languages, including the emotions
have a diverse dataset to work with. conveyed by emojis. By considering both the words
people use and the emojis they include, we can get a more
Preprocessing: We clean and prepare the text data, accurate picture of their feelings, even when they're using
removing any unnecessary elements and standardizing different languages.
the format to make it easier for the computer to
understand. Our work has important implications for understanding
people's emotions in the digital age, where
communication happens across languages and cultures. https://scholarworks.uark.edu/cgi/viewcontent.c
By improving how computers understand these emotions, gi?article=1078&context=csceuht
we can better analyze social media trends, customer
feedback, and public opinion, leading to more effective ➢ Guntakarla, D., & Zhou, M. (2018). A deep-
communication and decision-making. learning based framework for multi-lingual
sentiment analysis. Proceedings of the 2018
VII. REFERENCES Conference on Empirical Methods in Natural
Language Processing, 4908-4913.
➢ Pang, B., Lee, L., & Vaithyanathan, S. (2002). https://arxiv.org/abs/2206.12027
Thumbs up!: sentiment analysis using machine
learning techniques. ACL '02 Proceedings of the ➢ Bertran-Martinez, V., & Moreno, A. (2016).
40th Annual Meeting on Association for Sentiment analysis in multiple languages: A
Computational Linguistics, 793-803. comparative study. Information Processing &
https://www.cs.cornell.edu/home/llee/papers/sen Management, 52(2), 248-261.
timent.pdf https://pubmed.ncbi.nlm.nih.gov/37809397/

➢ Liu, B. (2010). Sentiment analysis and ➢ Xu, H., & Li, S. (2012). Word sentiment
subjectivity. In C. Manning & M. Jurafsky orientation and opinion intensity for social media
(Eds.), The handbook of natural language mining. Journal of Computational Information
processing (pp. 527-675). Cambridge University Systems, 7(3), 814-819.
https://www.researchgate.net/publication/22866 https://www.researchgate.net/publication/33523
7268_Sentiment_analysis_and_subjectivity 9475_Opinion_Mining_on_Social_Media_Data
_Sentiment_Analysis_of_User_Preferences
➢ Nakov, P., Ritter, A., & Rosenthal, S. (2016).
Semeval-2016 task 4: Sentiment analysis in ➢ Zadeh-Kalantari, S., Farzmand, H., & Samani, K.
twitter. Proceedings of the 10th International (2020). Multi-lingual sentiment analysis using
Workshop on Semantic Evaluation, 1-8. multi-lingual word embeddings and domain
https://arxiv.org/abs/1912.00741 adaptation. Expert Systems with Applications,
https://www.sciencedirect.com/science/article/pi
➢ Tang, D., Wei, F., Sun, S., & Wang, R. (2016). i/S0167739X19309719/
Emotional analysis of twitter text using emoji and
semantic lexicon. IEEE Transactions on
Computational Social Networks,
https://ieeexplore.ieee.org/document/8560896

➢ Maynard, D., & Ananiadou, S. (2012). Sentiment

analysis in multicultural contexts: The problem
of cultural relativism. Proceedings of the 40th
Annual Meeting on Association for
Computational Linguistics, 1320-1329.
https://www.um.edu.mt/library/oar/handle/1234
56789/46386

➢ Johansson, P., & Moschitti, A. (2013). Sentiment

analysis on twitter: A comparison of sentiment
lexicons. Proceedings of the 2013 Conference on
Empirical Methods in Natural Language
Processing, 5061-5073.