Text-Based Human Emotion Recognition Using Machine Learning Approach


Sentiment evaluation is a technique to pick out people’s attitudes, sentiments, and feelings in the direction of a given goal, such as people, activities, organizations, services, subjects, and products. Emotion detection is a subset of sentiment evaluation as it predicts the special emotion alternatively than simply declaring positive, negative, or neutral. In current times, many researchers have already labored on speech and facial expressions for text based human emotion recognition. However, emotion detection in textual content is a tedious challenge as cues are missing, in contrast to in speech, such as tonal stress, facial expression, pitch, etc. To discover thoughts from text, numerous techniques have been proposed in the previous the use of herbal language processing (NLP) techniques: the key-word approach, the lexicon-based approach, and the computing device gaining knowledge of approach. However, there have been some obstacles with keyword- and lexicon-based techniques as they center of attention on semantic relations. In this article, we have proposed a hybrid (machine learning + deep learning) mannequin to discover thoughts in text. Convolutional neural community (CNN) and Bi-GRU had been exploited as deep gaining knowledge of techniques. The support vector laptop is used as a desktop studying approach. The overall performance of the proposed strategy has evaluated the use of a mixture of three exceptional kinds of datasets, namely, sentences, tweets, and dialogs, and it attains an accuracy of 80.11%.


There have been many boundaries in the current solutions, such as that they did no longer have a listing of all the emotions. The current lists have an insufficient vocabulary of phrases in the lexicon, not noted words, semantics-based context, low extractions of contextual records from the given sentences, do no longer operate properly for detecting some unique emotions; susceptible context records extraction, unfastened semantic function extraction, much less computational speed, neglected family members between features, an insufficient quantity of data, and a excessive variety of misclassification. Some fashions had been no longer appropriate nicely for often going on emojis, vulnerable semantic statistics extraction, and the shape of the sentence. It differs from mannequin to model. There had been many boundaries in this device that had been fulfilled via preceding researchers. The proposed mannequin has fulfilled many of the present limitations.

Emotion detection is one of the large blessings of human-machine interplay as a nonliving aspect can experience or experience like a human being. Our proposed mannequin can discover feelings from textual content sentences that are tasteless as they do now not have any tone or expression. Many researchers have labored on a single dataset. But we have labored on three datasets which consist of the textual structure of simple sentences, tweets, and dialogs to notice emotions. Our text based human emotion recognition, awareness mannequin can be applied on any system. For commercial enterprise potential, this mannequin can assist to locate thoughts from purchaser reviews, services, supply protection for social media users, and many others.

The relaxation of the article is geared up as follows: Section two affords the literature survey on emotion detection. The proposed scheme is defined in Section three Section four attracts outcomes and evaluation of the proposed work. Finally, Section 5 concludes the article with future directions.

Literature Review

Several research have used a number strategies to discover thoughts from textual content. It will exhibit that which is the fantastic mannequin and offers us a greater accuracy.

Seal et al.  have carried out emotion detection with a keyword-based strategy mostly targeted on phrasal verbs. They used ISEAR data, preprocessed the data, and then utilized the keyword-based approach. They located various phrasal verbs that ought to have been related with emotion phrases however had been not, and so they constructed their very own database. They diagnosed phrasal verbs and key phrases synonymous with quite a number thoughts and categorized them the use of their database. They did, however, reap an awful lot greater accuracy of 65%, however they had been unable to tackle the researcher’s present issues, such as an inadequate listing of emotion key phrases and a lack of admire for phrase semantics in meaning. The work through Alotaibi has labored on a learning-based approach. He has used the ISEAR database for emotion detection. Then, the use of classifiers like Logistic Regression, K-Nearest Neighbour (KNN), XG-Boost, and Support Vector Machine (SVM), he preprocessed and skilled the data. According to him, all different classifiers poorly carried out as in contrast to Logistic Regression. Finally, he stated that the deep getting to know method would assist to enhance the model.

Proposed Scheme

This part describes the proposed device that consists of statistics series and information preprocessing. After preprocessing, the statistics will pass ahead as an enter to each ML and DL models. In ML, the preprocessed statistics will be given as the enter to ML classifiers and will exhibit the outcomes of all ML classifiers. Furthermore, it will choose the first-rate ML mannequin that offers the easiest accuracy. In DL, the statistics is transformed into vector shape and given as an enter to the DL models. Before that, we used a pretrained phrase vector to make the phrase embedding matrix and add the embedded layer to the DL mannequin [18]. After performing on person models, we mix the two fine DL fashions based totally on accuracy and F1 score. Combining them will supply us the latent vector and it will be given as an enter to the pleasant ML mannequin for the prediction of emotions. Finally, it will pick the quality accuracy of all the ML, DL, and hybrid models.

Results and Discussion

We have carried out many experiments the use of quite a number techniques to get the excellent accuracy for our proposed model. Emotion classification with a computing device getting to know approach, a deep gaining knowledge of approach, and our hybrid mannequin strategy on the multitext dataset consisting of sentences, tweets, and dialogs. Three datasets are used for performing these experiments.

Conclusion and Future Direction

In the future, we will attempt greater conceivable classifiers or ensemble strategies to enhance the results. In the deep mastering approach, we may also get the mixture of CNN, Bi-GRU, and LSTM to enhance the results. Additionally, we will work on the shape of textual content sentences and some of the regional languages. Moreover, in this digital world, people’s utilization of sending textual content messages, importing tweets, and writing on line opinions of merchandise have been in tremendous use and demand. Therefore, by way of having a lot of data, we can make a real-time text based human emotion recognition focus mannequin to discover the thoughts or moods of the people.

Data Availability

The records that guide the findings of this learn about are handy on request from the corresponding author.

What is text based human emotion recognition?

Emotion Detection in textual content files is in actuality a content material – based totally classification trouble involving standards from the domains of Natural Language Processing as nicely as Machine Learning. In this paper emotion focus based totally on textual information and the strategies used in emotion detection are discussed.

Which model is best for emotion recognition?

It was once concluded that the CNN model or CNN-LSTM hybrid models have been greater positive in emotional classification than regular computer mastering methods. In particular, the robotically function extraction of EEG indicators used to be validated to have excessive overall performance in end-to-end multi-dimensional emotion recognition.

Which algorithm is used in emotion detection?

Three famous ML algorithms, SVM, RF, and kNN have been used for emotion intensity recognition. A comparative learn about and implementation of algorithms for measuring facial thoughts and their intensities primarily based on the one-of-a-kind AUs (Action Units) are presented.

What are the 3 models of emotional intelligence?

There are three principal branches of emotional brain – the ability model, the trait model and the mixed model. There are three main branches of emotional intelligence – the potential model, the trait mannequin and the combined model.

What is human emotion recognition?

text based human emotion recognition is the procedure of figuring out human emotion. People fluctuate broadly in their accuracy at recognizing the thoughts of others. Use of science to assist humans with emotion, attention is an incredibly nascent lookup area. Generally, the science works pleasant if it makes use of more than one modality in context.

Please Visit for more project ideas: Project Ideas

if you want this project, please contact Pixel E-Services: Pixel E-Services

Leave a Reply