Search

Transfer learning models have proven superior to classical machine learning approaches in various text classification tasks, such as sentiment analysis, question answering, news categorization, and natural language inference. Recently, these models have shown exceptional results in natural language understanding (NLU). Advanced attention‐based language models like BERT and XLNet excel at handling complex tasks across diverse contexts. However, they encounter difficulties when applied to specific domains. Platforms like Facebook, characterized by continually evolving casual and sophisticated language, demand meticulous context analysis even from human users. The literature has proposed numerous solutions using statistical and machine learning techniques to predict the sentiment (positive or negative) of online customer reviews, but most of them rely on various business, review, and reviewer features, which leads to generalizability issues. Furthermore, there have been very few studies investigating the effectiveness of state‐of‐the‐art pre‐trained language models for sentiment classification in reviews. Therefore, this study aims to assess the effectiveness of BERT, RoBERTa, ALBERT, DistilBERT, and XLNet in sentiment classification using the Yelp reviews dataset. The models were fine‐tuned, and the results obtained with the same hyperparameters are as follows: 98.30 for RoBERTa, 98.20 for XLNet, 97.40 for BERT, 97.20 for ALBERT, and 96.00 for DistilBERT. [ABSTRACT FROM AUTHOR]

Aspect-level sentiment analysis is a critical component of sentiment analysis, aiming to determine the sentiment polarity associated with specific aspect words. However, existing methodologies have limitations in effectively managing aspect-level sentiment analysis. These limitations include insufficient utilization of syntactic information and an inability to precisely capture the contextual nuances surrounding aspect words. To address these issues, we propose an Aspect-Oriented Graph Attention Network (AOGAT) model. This model incorporates syntactic information to generate dynamic word vectors through the pre-trained model ALBERT and combines a graph attention network with BiGRU to capture both syntactic and semantic features. Additionally, the model introduces an aspect-focused attention mechanism to retrieve features related to aspect words and integrates the generated representations for sentiment classification. Our experiments on three datasets demonstrate that the AOGAT model outperforms traditional models. [ABSTRACT FROM AUTHOR]

In recent years, numerous achievements have been made in the field of deep learning, particularly in text processing. In the wave of intelligence, people's demand for intelligent communication is becoming increasingly higher. Therefore, we consider utilizing deep learning models to design and optimize transceiver of semantic communication system. The research of semantic communication is in a booming stage, but there are still few applications in multi-user scenario. In general, the parameters of the semantic communication system transceiver based on the deep learning model are very large. Therefore, we study the multi-user semantic communication system based on the ALBERT model. The goal of the proposed semantic communication system is to intelligently and correctly send the corresponding text classification to the receiver. The channel state information (CSI) is very important for information transmission. Considering the multi-antenna multi-user uplink scenario, we adopt the conditional generative adversarial network (cGAN) model to estimate CSI and apply it to the proposed semantic communication system. In order to reduce the influence of channel estimation on the delay of communication system, we quantify the pilot at the receiver. The simulation results show that the performance of the semantic communication system proposed in this paper is better than that of the semantic communication system based on Transformer model and the traditional semantic communication system in the intelligent text classification task. Moreover, in the case of low signal-to-noise ratio, traditional communication is difficult to complete intelligent tasks. [ABSTRACT FROM AUTHOR]

Cyber threat intelligence entity identification is the key to cyber threat intelligence analysis. In view of the fact that traditional word embedding cannot represent the polysemy of a word, it is difficult to effectively identify the key information of cyber threat intelligence entities, and at the same time, facing the exponential growth of threat intelligence, the efficiency of the identification model needs to be improved urgently. A network threat intelligence named entity recognition model based on ALBERT is proposed. The model first uses ALBERT to extract threat intelligence dynamic feature word vectors. Then, the feature word vector is input to the bidirectional long short-term memory network (BiLSTM) layer to obtain the corresponding label of each word in the sentence. Finally, the conditional random field (CRF) layer is modified and the sequence label is output with the maximum probability. The experimental results of identification model comparison show that the F1 value of the proposed model is 92.21%, which is obviously better than other models. In the case of the same recognition accuracy, the time and resource costs of the proposed model are also lower, which is suitable for massive and efficient entity recognition tasks in the field of cyber threat intelligence. [ABSTRACT FROM AUTHOR]

Named entity recognition can deeply explore semantic features and enhance the ability of vector representation of text data. This paper proposes a named entity recognition method based on multi-head attention to aim at the problem of fuzzy lexical boundary in Chinese named entity recognition. Firstly, Word2vec is used to extract word vectors, HMM is used to extract boundary vectors, ALBERT is used to extract character vectors, the Feedforward-attention mechanism is used to fuse the three vectors, and then the fused vectors representation is used to remove features by BiLSTM. Then multi-head attention is used to mine the potential word information in the text features. Finally, the text label classification results are output after the conditional random field screening. Through the verification of WeiboNER, MSRA, and CLUENER2020 datasets, the results show that the proposed algorithm can effectively improve the performance of named entity recognition. [ABSTRACT FROM AUTHOR]

Microblog is an important platform for mining public opinion, and it is of great value to conduct emotional analysis of microblog texts during the current epidemic. Aiming at the problem that most current emotional classification methods cannot effectively extract deep text features, and that traditional word vectors cannot dynamically obtain the semantics of words according to their context, which leads to classification bias, this research put forward a microblog text emotion classification algorithm based on TCN-BiGRU and dual attention (TCN-BiGRU-DATT). First, the vector representation of the text was obtained using ALBERT. Second, the TCN and BiGRU networks were used to extract the emotional information contained in the text through dual pathway feature extraction, to efficiently obtain the deep semantic features of the text. Then, the dual attention mechanism was introduced to allocate the global weight of the key information in the semantic features, and the emotional features were spliced and fused. Finally, the Softmax classifier was applied for emotion classification. The findings of a comparative experiment on a set of microblog text comments collected throughout the pandemic revealed that the accuracy, recall, and F1 value of the emotion classification method proposed in this paper reached 92.33%, 91.78%, and 91.52%, respectively, which was a significant improvement compared with other models. [ABSTRACT FROM AUTHOR]

This research aimed to evaluate the performance of the A Lite BERT (ALBERT), efficiently learning an encoder that classifies token replacements accurately (ELECTRA) and a robust optimized BERT pretraining approach (RoBERTa) models to support the development of the Indonesian language question and answer system model. The evaluation carried out used Indonesian, Malay and Esperanto. Here, Esperanto was used as a comparison of Indonesian because it is international, which does not belong to any person or country and this then make it neutral. Compared to other foreign languages, the structure and construction of Esperanto is relatively simple. The dataset used was the result of crawling Wikipedia for Indonesian and Open Super-large Crawled ALMAnaCH coRpus (OSCAR) for Esperanto. The size of the token dictionary used in the test used approximately 30,000 sub tokens in both the SentencePiece and byte-level byte pair encoding methods (ByteLevelBPE). The test was carried out with the learning rates of 1e-5 and 5e-5 for both languages in accordance with the reference from the bidirectional encoder representations from transformers (BERT) paper. As shown in the final result of this study, the ALBERT and RoBERTa models in Esperanto showed the results of the loss calculation that were not much different. This showed that the RoBERTa model was better to implement an Indonesian question and answer system. [ABSTRACT FROM AUTHOR]

Intent classification and named entity recognition of medical questions are two key subtasks of the natural language understanding module in the question answering system. Most existing methods usually treat medical queries intent classification and named entity recognition as two separate tasks, ignoring the close relationship between the two tasks. In order to optimize the effect of medical queries intent classification and named entity recognition tasks, a multi-task learning model based on ALBERT-BILSTM is proposed for intent classification and named entity recognition of Chinese online medical questions. The multi-task learning model in this paper makes use of encoder parameter sharing, which enables the model's underlying network to take into account both named entity recognition and intent classification features. The model learns the shared information between the two tasks while maintaining its unique characteristics during the decoding phase. The ALBERT pre-training language model is used to obtain word vectors containing semantic information and the bidirectional LSTM network is used for training. A comparative experiment of different models was conducted on Chinese medical questions dataset. Experimental results show that the proposed multi-task learning method outperforms the benchmark method in terms of precision, recall and F1 value. Compared with the single-task model, the generalization ability of the model has been improved. [ABSTRACT FROM AUTHOR]

With the world's interaction moving more and more toward using online social media platforms, the advent of cyberbullying has also raised its head. Multiple forms of cyberbullying exist from the more common text based to images or even videos, and this paper will explore the context of textual comments. Even in the niche area of considering only text-based data, several approaches have already been worked upon such as n-grams, recurrent units, convolutional neural networks (CNNs), gated recurrent unit (GRU) and even a combination of the mentioned architectures. While all of these produce workable results, the main point of contention is that true contextual understanding is quite a complex concept. These methods fail due to two simple reasons: (i) lack of large datasets to properly utilize these architectures and (ii) the fact that understanding context requires some mechanism of remembering history that is only present in the recurrent units. This paper explores some of the recent approaches to the difficulties of contextual understanding and proposes an ALBERT-based fine-tuned model that achieves state-of-the-art results. ALBERT is a transformer-based architecture and thus even at its untrained form provides better contextual understanding than other recurrent units. This coupled with the fact that ALBERT is pre-trained on a large corpus allowing the flexibility to use a smaller dataset for fine-tuning as the pre-trained model already has deep understanding of the complexities of the human language. ALBERT showcases high scores in multiple benchmarks such as the GLUE and SQuAD showing that high levels of contextual understanding are inherently present and thus fine-tuning for the specific case of cyberbullying allows to use this to our advantage. With this approach, we have achieved an F1 score of 95% which surpasses current approaches such as the CNN + wordVec, CNN + GRU and BERT implementations. [ABSTRACT FROM AUTHOR]

The upsurge of prolific blogging and microblogging platforms enabled the abusers to spread negativity and threats greater than ever. Negative and hateful comments are averting users from sharing their opinion freely on social media platforms. It often breaks people's confidence and causes extensive damage to their mental health. Hence, identifying these toxic contents and taking appropriate measures against them is crucial to preserve a safe environment on social media. Numerous state-of-the-art approaches classify the whole content as toxic or non-toxic, but they don't distinguish the precise toxic portion from the whole content. Detecting the toxic portions is essential as it substantially aids to moderate the toxic contents through excluding the abusive parts. This paper describes our proposed approach to detect the toxic portions from text contents efficiently and accurately. We explore an ensemble of sequence labeling models including the word embedding-based Spark NLP NER (named entity recognition) deep learning model, spaCy NER model with custom toxic tags, and ALBERT NER model to identify the toxic spans. The NER-based models usually intend to capture the contextual attributes of phrases and spans that are essential for named entity recognition. As the toxic span detection task also requires us to apprehend the phrasal context for detecting toxic span, the similarity between these two tasks inspires us to exploit these NER models. Finally, we determine the final toxic spans using a prevalence-based fusion of the predictions generated by these models. The fusion strategy enables us to consolidate the diversity of these models for perceiving the phrasal context in all aspects. Experimental results achieved on the SemEval-2021 toxic spans detection dataset depict that our model meticulously captures the toxic fragment and achieves a competitive result among the other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Using, as a point of reference, Albert Memmi's well-known description of the colonial situation as envisaged in 1957, I address the postcolonial treatment of Albert Camus. While criticisms of his writings on Algeria have been frequent, my argument is that an overly politicized reading of his work, especially his Algerian Chronicles, leaves aside, avoids, or misinterprets important and still-relevant ethical messages concerning violence, terrorism, and repression. In my rereading of Camus's truce appeals, I propose seeing Camus outside the colonizer-colonized dichotomy, as a person, above all, engaged in seeking the possibility of dialogue among the French, Arabs, and Berbers. While Camus's political vision, as has already been demonstrated, lacks sympathy for the insurgents' will for total independence from France, it contains a serious proposition to apply moral limits to the means of the struggle, consistent with Camus's philosophical deliberations from The Rebel. [ABSTRACT FROM AUTHOR]

This paper will address René Girard's critique of the "humanization of nothingness" in modern Western philosophy. I will first explain how the "desire for death" is related to a phenomenon that Girard refers to as "obstacle addiction." Second, I will point out how mankind's desire for death and illusory will to self-divinization gradually tend to converge within the history of modern Western humanism. In particular, I will show how this convergence between selfdestruction and self-divinization gradually takes shape through the evolution of the concept of "the negative" from Hegel to Kojève, Sartre and Camus. Finally, we shall come to see that in Girard's view "the negative" has tended to become an ever-preoccupying and unacknowledged symptom of mankind's addiction to "model/obstacles" of desire. [ABSTRACT FROM AUTHOR]

I take the belated and confused response to Edward Said’sOrientalism(1978) as due not only to the pretentious “verbosity” of the first translation, but also to objective circumstances of a political and ideological nature. Frantz Fanon, for instance, made a significant belated presence in the 1960s through translation and ultimate appeal to the rising critique of the national elite thereafter. Gramsci was already known among the Left. Not so is the case withOrientalism. The confusion surrounding the term, the distrust of the West, and the affiliation of prominent Arabists and Orientalists in Arabic language societies in Damascus and Cairo and their significant contributions to the study of Arabic added another dimension to a blurred picture. As noticed by some scholars, the much deferred application of Said’s contrapuntal cultural criticism rather than his postcolonial drive, speaks of a different politics. Largely but not solely indebted to Marxist theory, this liberatory politics in gender, race and class, and the significant interest in social equality, justice and political/economic and cultural liberation, take a number of directions that surpass and also obfuscate postcolonial experimentations and their parlance. To demonstrate the efficacy of this massive engagement in the devaluation of postcolonial discourse, I use among others three early texts of the 1930s that put us face to face with a postcolonial consciousness that not only exposes colonialism and the cultural dependency of the elite, and sets the ground for political awareness, but also brings forth the issues of biopolitics and the implications of subalternity along a number of axes. This line of thought is predicated on a cultural and ideological script that had already taken root in urban centres since the early twentieth century. Although the Cold War situation and the enforced assimilation of the Left into statist national blocs or confederations or the denial and prosecution of opposition intensified a polarized polemic against the “West”, there was already that script that was given legitimacy and power in national democratic writings. Hence, whatever is subsumed under the rubric of the “postcolonial” sounds like “carrying dates to Hajar” (ka-nāqil tamrin 'ilā Hajar). Put together in a genealogical map, these early engagements, including the forceful critique of the “West” and its body of representations, gain further solidification that derails and incapacitates the “postcolonial” as a viable category of analysis and application in the Arab world. [ABSTRACT FROM PUBLISHER]

In this paper, we aim at improving Japanese text classification using TextCNN-based ensemble learning model. Specifically, we first construct three different sub-classifiers, combining ALBERT, RoBERTa, DistilBERT with TextCNN, respectively; and then explore the effectiveness of ensemble learning model to leverage complementary information from different sub-classifiers for better text classification. We also conduct a series of experiments with the dataset collected from Japanese Wikipedia pages, which was divided into 31 categories. The experimental results show that the proposed approach achieves a good performance. The accuracy, precision, recall and F1 scores reach 0.881 , 0.884 , 0.880 and 0.881 , respectively, which shows that the TextCNN-based ensemble learning model can be used for Japanese Text Multi-Classification effectively. [Display omitted] • Three TextCNN-based sub-classifiers for Japanese text classification are designed. • A Bagging ensemble learning model is proposed to combine three different subclassifiers for multi-label Japanese text classification. • A Japanese dataset for classification is constructed by crawling from Japanese Wikipedia pages. • The TextCNN-based model with bagging ensemble learning performs well for text classification. [ABSTRACT FROM AUTHOR]

We observe several facts and make conjectures about commutative algebras satisfying the Jacobi identity. The central question is which of those algebras admit a faithful representation (i.e., in Lie parlance, satisfy the Ado theorem, or, in Jordan parlance, are special). [ABSTRACT FROM AUTHOR]

After absorbing and eventually halting the German spring offensive by July 1918, the Allies launched a major counter-offensive. On 8 August, 31 divisions -- 6 British, 5 Australian, 4 Canadian and 16 French -- and an American regiment, each supported by tanks, artillery and aircraft, began what would become series of attacks over several weeks in the Somme sector. Applying tactics that had been learned over years of war, they tore a great hole in the German lines and, by the end of August, had driven the German Army back to the Hindenburg Line -- regaining most of the ground that had been lost to the Germans in their spring offensive. More significantly, the counter-offensive caused the German Army's morale to collapse -- an end to the war seemed near. [ABSTRACT FROM AUTHOR]

The very dubious results of Der Positivismusstreit are explained as consequences of a methodological breakdown between two schools. But this debate is nevertheless relevant for today’s discussions due to its historical role in setting today’s framework. The methodological breakdown between the schools is overcome to a partial degree by presuming some overlap in methods. The misunderstandings of the past are set in today’s context, where critical realists have abandoned much of their previous doctrines, and the failure of critical rationalists to acquire more influence than they have is explained as in large part due to their failure to avoid identification with conservative economic-cum-political theory, above all that of Hayek. This mistake is corrected by explaining how improved critical rationalist methods are consistent with a variety of political alternatives and may be used to critically examine them. [ABSTRACT FROM PUBLISHER]

Abstract: Premature birth is a major and growing problem. Investigations into neuroanatomical correlates and consequences of preterm birth are hampered by complex neonatal brain anatomy and unavailability of atlases and protocols covering the whole brain. We developed delineation protocols for the manual segmentation of cerebral magnetic resonance (MR) images from newborn infants into 50 regions with comprehensive coverage of the brain. We then segmented MR scans from 15 infants born preterm at median 29, range 26–35, weeks postmenstrual age and scanned at term-corrected age, and five term-born infants born at median 41, range 39–45, weeks postmenstrual age. Total and regional brain volumes were estimated in each infant, and regional volumes expressed as a fraction of total brain volume. Total brain volumes were higher with greater age at birth and at time of scan, but once corrected for age at scan there was no difference between preterm and term infants. Fractional age-corrected regional volumes were bigger unilaterally in terms in middle and inferior temporal gyri, anterior temporal lobe, fusiform gyrus and posterior cingulate gyrus. Fractional age-corrected regional volumes were larger in preterms bilaterally in hippocampus, amygdala, thalamus and lateral ventricles, left superior temporal gyrus and right caudate nucleus. These differences were not significant after correcting for multiple hypothesis testing, but suggest subtle differences between preterms and term-borns accessible to regional analysis. Detailed illustrated protocols are made available in the Appendix. [Copyright &y& Elsevier]

I defend the idea that the fact that no system is entirely isolated ("Interventionism") can be used to explain the successful use of the microcanonical distribution in statistical mechanics. The argument turns on claims about what is needed for an adequate explanation of this fact: I argue in particular that various competing explanations do not meet reasonable conditions of adequacy, and that the most striking lacuna in Interventionism-its failure to explain the "arrow of time"-is no real defect. [ABSTRACT FROM AUTHOR]

From his obituary of Samuel Hammerschlag, we know of Freud's great veneration for his teacher of Jewish religion. However, not only Hammerschlag himself but his whole family had a formative influence on young Freud, who was deeply impressed by their humanity. This paper describes Freud's relationships with all the family members. In particular, it shows how warmly he felt towards the only daughter, Anna Hammerschlag, who was his patient for a while and whom he chose as a godmother for his youngest daughter Anna. By virtue of the crucial role she played in Freud's 'specimen dream' of July 1895 ('Irma's injection'), she also became as it were the godmother of Freud's magnum opus, The Interpretation of Dreams. All the known extant letters from Freud to members of the Hammerschlag family are published here for the first time in English translation. [ABSTRACT FROM AUTHOR]

Bovine Spongiform Encephalopathy (BSE), popularly known as ‘Mad Cow’ disease, was discovered in the late 1980s in Britain; in 1996, scientists announced a ‘probable’ link between eating BSE-contaminated meat and a new form of Creutzfeldt-Jakob disease, a fatal human brain disease. Britain's beef industry was devastated, beef consumption dropped, export markets closed and a mass cull of older cattle was implemented. This article uses social representation theory to analyze how Canada's 2003 BSE outbreak was portrayed to Canadians in major newspapers and compares this representation with BSE's impact in two rural Alberta counties. The day Canada's BSE case was reported, the United States closed its border to Canadian cattle and beef. The event was represented as ‘devastating’ to Canada's cattle and beef industries and rural areas in general, a view that went largely unchallenged and was critical to gaining government support for the affected industries. Little evidence of economic devastation was found in the heart of Alberta's cow-calf producing area; producers adapted to their changing economic circumstances and acquired other sources of income. But there is little doubt that the financial uncertainty associated with BSE added to stress levels among farm families. [ABSTRACT FROM AUTHOR]

It is misleading to speak of warfare between science and Christian theology, as Andrew White did in 1896. White also was mistaken in exaggerating the conflict between the church and Galileo and Copernicus. The more important issue between science and theology has to do with the mechanistic interpretation of nature. When he introduced the principle of inertia in his natural philosophy, René Descartes insisted that God's immutability renders it impossible for God to intervene in the creation. He reduced the idea of God to a deistic notion by speaking of motion exclusively as a property of bodies. Even though Isaac Newton offered a different view, the Cartesian view dominated subsequent thinking. This made dialogue with theology difficult. Michael Faraday, followed by Albert Einstein, introduced the idea of field; bodily phenomena were subordinated as manifestations of fields. The precursor of the idea of field is the Stoic idea of spirit, which is close to the biblical concept of spirit. Thomas Torrance and I have taken this concept of field as an occasion to reopen dialogue. Mechanistic thinking accounts for the tension between Darwinian thought and theology. In principle the tension can be resolved, because the Bible itself asserts that all living things were brought from the earth—that is, organic life emerged from inorganic matter. Thus, emergence, contingency, and novelty are consistent with Darwinian evolutionary thinking. Contingency can be related conceptually to the activity of God in creation. [ABSTRACT FROM AUTHOR]

Deep learning is advancing rapidly; it has aided in solving problems that were thought impossible. Natural language understanding is one such task that has evolved with the advancement of deep learning systems. There have been several sentiment analysis attempts, but they aim to classify it as a single emotion. Human emotion in natural language is generally a complex combination of emotions, which may be indeterminate or neutral at times. Neutrosophy is a branch of philosophy that identifies neutralities and uses membership functions (positive, negative, neutral) to quantify a sample into Single Valued Neutrosophic Set (SVNS) values. Our work aims to combine the power of deep learning with SVNS to represent a sample's sentiment into membership functions of SVNS. We have worked on the Offensive Language Identification Dataset (OLID). Combining the power of state-of-the-art neural network techniques with neutrosophy allowed us to quantify the sentiments and identify the transition phase between positive and negative ones. We used the transition phase to capture neutral samples, which is beneficial if we want to obtain purely positive/negative samples. We performed experiments using Bi-directional Long Short Term Memory (BiLSTM) with attention, Bidirectional Encoder Representations from Transformers (BERT), A Lite BERT (ALBERT), A Robustly Optimised BERT Approach (RoBERTa), and MPNet. Our SVNS model performed equivalent to state-of-the-art neural network models on the OLID dataset. Here, we propose a novel framework that can integrate with any neural network model and quantify sentiments using SVNS. • Sentiment analysis model using deep learning, neutrosophy, and transfer learning. • Analysing tweets as a combination of sentiments identifying neutralities. • Quantifying tweets into Single Valued Neutrosophic Sets (SVNS) for OLID dataset. • Experimental analysis using BiLSTM, BERT, RoBERTa, ALBERT, and MPNet. • Gaussian Mixture Model and k-means clustering algorithm for SVNS calculation. [ABSTRACT FROM AUTHOR]

This paper introduces a series of experiments with an ALBERT over match-LSTM network on the top of pre-trained word vectors, for accurate classification of intelligent question answering and thus the guarantee of precise information service. To improve the performance of data classification, a short text classification method based on an ALBERT and match-LSTM model was proposed to overcome the limitations of the classification process, such as few vocabularies, sparse features, large amount of data, lots of noise and poor normalization. In the model, Jieba word segmentation tools and agricultural dictionary were selected to text segmentation, GloVe algorithm was then adopted to expand the text characteristic and weighted word vector according to the text of key vector, bi-directional gated recurrent unit was applied to catch the context feature information and multi-convolutional neural networks were finally established to gain local multidimensional characteristics of text. Batch normalization, Dropout, Global Average Pooling and Global Max Pooling were utilized to solve overfitting problem. The results showed that the model could classify questions accurately, with a precision of 96.8%. Compared with other classification models, such as multi-SVM model and CNN model, ALBERT+match-LSTM had obvious advantages in classification performance in intelligent Agri-tech information service. [ABSTRACT FROM AUTHOR]

Deep learning has been widely used in numerous real-world engineering applications and for classification problems. Real-world data is present with neutrality and indeterminacy, which neutrosophic theory captures clearly. Though both are currently developing research areas, there has been little study on their interlinking. We have proposed a novel framework to implement neutrosophy in deep learning models. Instead of just predicting a single class as output, we have quantified the sentiments using three membership functions to understand them better. Our proposed model consists of two blocks, feature extraction, and feature classification. Having a separate feature extraction block enables us to use any model as a feature extractor. We experimented with BiLSTM using GloVe (Global Vectors for word representation), BERT (Bidirectional Encoder Representations from Transformers), ALBERT (A Lite BERT), RoBERTa (Robustly optimized BERT approach), MPNet, and stacked ensemble models. Feature classification performs prediction and dimensionality reduction of features. Experimental analysis was done on the SemEval 2017 Task 4 dataset (Subtask A). We used the intermediate layer features to define membership functions of Single Valued Neutrosophic Sets (SVNS). We used these membership functions for prediction as well. We have compared our models with the top five teams of the task and recent state-of-the-art systems. Our proposed stacked ensemble model achieved the best recall (0.733) score. • Architecture for combining neutrosophy and deep learning for sentiment analysis. • Experimental analysis using LSTM, BERT, RoBERTa, ALBERT, MPNet, and Ensemble models. • Feature extraction using intermediate layers of deep learning models. • Using K-means and Cosine distance for calculating SVNS from generated features. • SVNS as an alternative to Softmax for multi-label sentiment analysis problem. [ABSTRACT FROM AUTHOR]