Sentiment analysis of the Hamas-Israel war on YouTube comments using deep learning Scientific Reports

The process of converting preprocessed textual data to a format that the machine can understand is called word representation or text vectorization. One can train machines to make near-accurate predictions by providing text samples as input to semantically-enhanced ML algorithms. Machine learning-based semantic analysis involves sub-tasks such as relationship extraction and word sense disambiguation. Prediction-based embeddings are word representations derived from models that are trained to predict certain aspects of a word’s context or neighboring words.

Moreover, when support agents interact with customers, they are able to adapt their conversation based on the customers’ emotional state which typical NLP models neglect. Therefore, startups are creating NLP models that understand the emotional or sentimental aspect of text data along with its context. Such NLP models improve customer loyalty and retention by delivering better services and customer experiences. The simple default classifier I’ll use to compare performances of different datasets will be the logistic regression.

The models utilized in this study were constructed using various algorithms, incorporating the optimal parameters for each algorithm. The evaluation of model performance was based on several metrics, including accuracy, precision, recall, and F1. Accuracy, precision, recall, and F1 are commonly employed to assess the performance of classification models.

Why We Picked SAP HANA Sentiment Analysis

These models excel in complex sentences with multiple aspects, adjusting focus to relevant segments and improving sentiment predictions. Their interpretability and enhanced performance across various ABSA tasks underscore their significance in the field65,66,67. Created by Facebook’s AI research team, the library enables you to carry out many different applications, including sentiment analysis, where it can detect if a sentence is positive or negative. VADER calculates the text sentiment and returns the probability of a given input sentence to be positive, negative, or neural.

One potential solution to address the challenge of inaccurate translations entails leveraging human translation or a hybrid approach that combines machine and human translation. Human translation offers a more nuanced and precise rendition of the source text by considering contextual factors, idiomatic expressions, and cultural disparities that machine translation may overlook. However, it is essential to note that this approach can be resource-intensive in terms of time and cost. Nevertheless, its adoption can yield heightened accuracy, especially in specific applications that require meticulous linguistic analysis. The machine learning model is trained to analyze topics under regular social media feeds, posts and revews.

Published in Towards Data Science

If you want to know more about precision and recall, you can check my old post, “Another Twitter sentiment analysis with Python — Part4”. The data is not well balanced, and negative class has the least number of data entries with 6,485, and the neutral class has the most data with 19,466 entries. I want to rebalance the data so that I will have a balanced dataset at least for training.

Based on character level features, the one layer CNN, Bi-LSTM, twenty-nine layers CNN, GRU, and Bi-GRU achieved the best measures consecutively. A sentiment categorization model that employed a sentiment lexicon, CNN, and Bi-GRU was proposed in38. Sentiment weights calculated from the sentiment lexicon were used to weigh the input embedding vectors. The CNN-Bi-GRU network detected both sentiment and context features from product reviews better than the networks that applied only CNN or Bi-GRU. Meanwhile, many customers create and share content about their experience on review sites, social channels, blogs etc. The valuable information in the authors tweets, reviews, comments, posts, and form submissions stimulated the necessity of manipulating this massive data.

Thus, I investigated the discrepancies and gave my ruling, to which either Humans or the Chatgpt I found was more precise. The final result is displayed in the plot below, which shows how the accuracy (y-axis) changes for both models when categorizing the numeric Gold-Standard dataset, as the threshold (x-axis) is adjusted. Also, the training and testing sets are on the left and right sides, respectively. Still, as an AI researcher, industry professional, and hobbyist, I am used to fine-tuning general domain NLP machine learning tools (e.g., GloVe) for usage in domain-specific tasks. This is the case because it was uncommon for most domains to find an out-of-the-box solution that could do well enough without some fine-tuning. Employee sentiment analysis enables HR to more easily and effectively obtain useful insights about what employees think about the organization by analyzing how they communicate in their work environment.

• The demo program loads the training data into a meta-list using a specific format that is required by the EmbeddingBag class.
• The extra dimension that wasn’t available to us in our original matrix, the r dimension, is the amount of latent concepts.
• Additionally, the solution integrates with a wide range of apps and processes as well as provides an application programming interface (API) for special integrations.

The standard CNN structure is composed of a convolutional layer and a pooling layer, followed by a fully-connected layer. Some studies122,123,124,125,126,127 utilized standard CNN to construct classification models, and combined other features such as LIWC, TF-IDF, BOW, and POS. In order to capture sentiment information, Rao et al. proposed a hierarchical MGL-CNN model based on CNN128. Lin et al. designed a CNN framework combined with a graph model to leverage tweet content and social interaction information129.

Large volumes of data can be analyzed by deep learning algorithms, which can identify intricate relationships and patterns that conventional machine learning methods might overlook20. The context of the YouTube comments, including the author’s location, demographics, and political affiliation, can also be analyzed using deep learning techniques. In this study, the researcher has successfully implemented a deep neural network with seven layers of movie review data. The proposed model achieves an accuracy of 91.18%, recall of 92.53%, F1-Score of 91.94%, and precision of 91.79%21. Today, semantic analysis methods are extensively used by language translators. Earlier, tools such as Google translate were suitable for word-to-word translations.

According to its documentation, it supports sentiment analysis for 136 languages. Polyglot is often chosen for projects that involve languages not supported by spaCy. Idiomatic has recently introduced its granularity generator feature, which reads tickets, summarizes key themes, and finds sub-granular issues to get a more holistic context of customer feedback. It also developed an evaluating chatbot performance feature, which offers a data-driven approach to a chatbot’s effectiveness so you can discover which workflows or questions bring in more conversions. Additionally, Idiomatic has added a sentiment score tool that calculates the score per ticket and shows the average score per issue, desk channel, and customer segment.

The revealed information is an essential requirement to make informed business decisions. Understanding individuals sentiment is the basis of understanding, predicting, and directing their behaviours. By applying NLP techniques, SA detects the polarity of the opinioned text and classifies it according to a set of predefined classes. NLP tasks were investigated by applying statistical and machine learning techniques. Deep learning models can identify and learn features from raw data, and they registered superior performance in various fields12.

In the fourth phase of the methodology, we conducted sentiment analysis on the translated data using pre-trained sentiment analysis deep learning models and the proposed ensemble model. The ensemble sentiment analysis model analyzed the text to determine the sentiment polarity (positive, negative, or neutral). The algorithm shows step by step process followed in the sentiment analysis phase.

NLTK also provides access to more than 50 corpora (large collections of text) and lexicons for use in natural language processing projects. Another potential challenge in translating foreign language text for sentiment analysis is irony or sarcasm, which can prove intricate in identifying and interpreting, even for native speakers. Irony and sarcasm involve using language to express the opposite of the intended meaning, often for humorous purposes47,48. For instance, a French review may use irony or sarcasm to convey a negative sentiment; however, individuals lacking fluency in French may struggle to comprehend this intended tone. You can foun additiona information about ai customer service and artificial intelligence and NLP. Similarly, a social media post in German may employ irony or sarcasm to express a positive sentiment, but this could be arduous to discern for those unfamiliar with language and culture.

Data

Most implementations of LSTMs and GRUs for Arabic SA employed word embedding to encode words by real value vectors. Besides, the common CNN-LSTM combination applied for Arabic SA used only one convolutional layer and one LSTM layer. Contrary to RNN, gated variants are capable of handling long term dependencies. Also, they can combat vanishing and exploding gradients by the gating technique14. Bi-directional recurrent networks can handle the case when the output is predicted based on the input sequence’s surrounding components18. LSTM is the most widespread DL architecture applied to NLP as it can capture far distance dependency of terms15.

Now that we’ve selected our architecture from an initial search of XGBoost, LGBM and a simple keras implementation of a neural network, we’ll need to conduct a hyperparameter optimization to fine-tune our model. Hyperparameter optimization can be an incredibly difficult, computationally expensive, and slow process for complicating modeling tasks. Comet has built an optimization service that can conduct this search for you. Simply pass in the algorithm you’d like to sweep the hyperparameter space with, hyperparameters and ranges to search, and a metric to minimize or maximize, and Comet can handle this part of your modeling process for you. Next, we’ll build a Light Gradient-Boosting classifier (LGBM), an XGBoost classifier, and a relatively straightforward neural network with keras and compare how each of these models performs. Oftentimes it’s hard to tell which architecture will perform best without testing them out.

We will be talking specifically about the English language syntax and structure in this section. Considering a sentence, “The brown fox is quick and he is jumping over the lazy dog”, it is made of a bunch of words and just looking at the words by themselves don’t tell us much. We, now, have a neatly formatted dataset of news articles and you can quickly check the total number of news articles with the following code. We will now build a function which will leverage requests to access and get the HTML content from the landing pages of each of the three news categories. Then, we will use BeautifulSoup to parse and extract the news headline and article textual content for all the news articles in each category. We find the content by accessing the specific HTML tags and classes, where they are present (a sample of which I depicted in the previous figure).

These challenges necessitate ongoing research and development of more sophisticated ABSA models that can navigate the intricacies of sentiment analysis with greater accuracy and contextual sensitivity. The Python library can help you carry out sentiment analysis to analyze opinions or feelings through data by training a model that can output if text is positive or negative. It provides several vectorizers to translate the input documents into vectors of features, and it comes with a number of different classifiers already built-in. The simple Python library supports complex analysis and operations on textual data. For lexicon-based approaches, TextBlob defines a sentiment by its semantic orientation and the intensity of each word in a sentence, which requires a pre-defined dictionary classifying negative and positive words. The tool assigns individual scores to all the words, and a final sentiment is calculated.

Its documentation shows that it supports tokenization for 165 languages, language detection for 196 languages, and part-of-speech tagging for 16 languages. With its intuitive interfaces, Gensim achieves efficient multicore implementations of algorithms like Latent Semantic Analysis (LSA) and Latent Dirichlet Allocation (LDA). Some of the library’s other top use cases include finding text similarity and converting words and documents to vectors.

In the chart below we can see the distrubution of polarity on a scale -1 to 1 for customer reviews based on recommendations. The algorithm forms a prediction based on the current behavioral pattern of the anomaly. If the predicted values exceed the threshold confirmed during the training phase, an alert is sent.

Its interface also features a properties panel, which lets you select a target variable, and advanced panels to select languages, media types, the option to report profanities, and more. IBM Watson NLU recently announced the general availability of a new single-label text classification capability. This new feature extends language support and enhances training data customization, suited for building a custom sentiment classifier. Once the model is trained, it will be automatically deployed on the NLU platform and can be used for analyzing calls. The following table provides an at-a-glance summary of the essential features and pricing plans of the top sentiment analysis tools. Finally, we applied three different text vectorization techniques, FastText, Word2vec, and GloVe, to the cleaned dataset obtained after finishing the preprocessing steps.

Automated ticketing support

Platform limits, as well as data bias, have the potential to compromise the dataset’s trustworthiness and representativeness. Furthermore, the sheer volume of comments and the dynamic nature of online discourse may necessitate scalable and effective data collection and processing approaches. Next, the experiments were accompanied by changing different hyperparameters until we obtained a better-performing model in support of previous works. During the experimentation, we used techniques like Early-stopping, and Dropout to prevent overfitting. The models used in this experiment were LSTM, GRU, Bi-LSTM, and CNN-Bi-LSTM with Word2vec, GloVe, and FastText. This study was used to visualize YouTube users’ trends from the proposed class perspectives and to visualize the model training history.

Translators often face challenges in rendering core concepts into alternative words or phrases while striving to maintain fidelity to the original text. Yet, even with the translators’ understanding of these core concepts, significant variations emerge in their specific word choices. These variations, along with the high frequency of core concepts in the translations, directly contribute ChatGPT App to differences in semantic representation across different translations. The translation of The Analects contains several common words, often referred to as “stop words” in the field of Natural Language Processing (NLP). These words, such as “the,” “to,” “of,” “is,” “and,” and “be,” are typically filtered out during data pre-processing due to their high frequency and low semantic weight.

Rules are established on a comment level with individual words given a positive or negative score. If the total number of positive words exceeds negative words, the text might be given a positive sentiment and vice versa. SpaCy supports more than 75 languages and offers 84 trained pipelines for 25 of these languages. It also integrates with modern transformer models like BERT, adding even more flexibility for advanced NLP applications. By highlighting these contributions, this study demonstrates the novel aspects of this research and its potential impact on sentiment analysis and language translation. Last on our list is PyNLPl (Pineapple), a Python library that is made of several custom Python modules designed specifically for NLP tasks.

The Hummingbird algorithm was formed in 2013 and helps analyze user intentions as and when they use the google search engine. As a result of Hummingbird, results are shortlisted based on the ‘semantic’ relevance of the keywords. Moreover, it also plays a crucial role in offering SEO benefits to the company.

For example, in several of my NLP projects I wanted to retain the word « don’t » rather than split it into three separate tokens. One approach to create a custom tokenizer is to refactor the TorchText basic_english tokenizer source code. The MyTokenizer class constructs a regular expression and the tokenize() method applies the regular expression to its input text.

Supporting the GRU model with handcrafted features about time, content, and user boosted the recall measure. A comparative study was conducted applying multiple deep learning models based on word and character features37. Three CNN and five RNN networks were implemented and compared on thirteen reviews datasets. Although the thirteen datasets included reviews, the deep models performance varied according to the domain and the characteristics of the dataset. Based on word-level features Bi-LSTM, GRU, Bi-GRU, and the one layer CNN reached the highest performance on numerous review sets, respectively.

Microsoft Azure AI Language (formerly Azure Cognitive Service for Language) is a cloud-based service that provides natural language processing (NLP) features and is designed to help businesses harness the power of textual data. It offers semantic analysis nlp a wide range of capabilities, including sentiment analysis, key phrase extraction, entity recognition, and topic moderation. Azure AI Language translates more than 100 languages and dialects, including some deemed at-risk and endangered.

This gives significant insight for spam and fraudulent news and posts detection. Getting started with GPT-4 involves setting up the necessary software and hardware environment, obtaining the model, and learning how to use it. There are various resources available online, including tutorials, documentation, and community forums, that can help you get started. You will also need a suitable dataset for training or fine-tuning the model, depending on your specific use case. This allows it to understand the syntax and semantics of various programming languages.

RNNs capture the pattern in the time dimension, while convolutional neural networks (CNN) capture the pattern in the space dimension. CNN works well in long-range semantic comprehension and detects the local and position-defined pattern. The model generates a feature map of sentences by using k-max-pooling to obtain the short and long relationship between words and phrases.

• Instead of treating all parts of the input equally, attention mechanisms allow the model to selectively attend to relevant portions of the input.
• The plot below shows bimodal distributions in both training and testing sets.
• The following two interactive plots let you explore the reviews by hovering over them.
• It is nearly impossible to study Confucius’s thought without becoming familiar with a few core concepts (LaFleur, 2016), comprehending the meaning is a prerequisite for readers.
• NLP algorithms generate summaries by paraphrasing the content so it differs from the original text but contains all essential information.

Three sarcasm identification corpora containing tweets, quote responses, news headlines were used for evaluation. The proposed representation integrated word embedding, weighting functions, and N-gram techniques. The weighted representation of a document was computed as the concatenation of the weighted unigram, bigram and trigram representations.

By combining both LSTM and GRU in an ensemble model, the objective is to enhance long-term dependency modelling and improve accuracy. The ensemble model consists of an LSTM layer followed by a GRU layer, where the output from LSTM serves as input for GRU. Aslam et al. (2022) performed sentiment analysis and emotion detection on tweets related to cryptocurrency. TextBlob libraries are used to annotate sentiment, and Text2emotion is used to detect emotions such as angry, fear, happy, sad and surprise. They use different settings of feature extraction, which are Bag-of-word, TF-IDF and Word2Vec. They build several machine learning classifiers and deep learning classifiers using the neural network LSTM and GRU.

SemEval (Semantic Evaluation) is a renowned NLP workshop where research teams compete scientifically in sentiment analysis, text similarity, and question-answering tasks. The organizers provide textual data and gold-standard datasets created by annotators (domain specialists) and linguists to evaluate state-of-the-art solutions for each task. There has been growing research interest in the detection of mental illness from text. Early detection of mental disorders is an important and effective way to improve mental health diagnosis.

Topic modeling helps in exploring large amounts of text data, finding clusters of words, similarity between documents, and discovering abstract topics. As if these reasons weren’t compelling enough, topic modeling is also used in search engines wherein the search string is matched with the results. Non-negative matrix factorization ChatGPT (NMF )can be applied for topic modeling, where the input is term-document matrix, typically TF-IDF normalized. It is derived from multivariate analysis and linear algebra where a matrix Ais factorized into (usually) two matrices W and H, with the property that all three matrices have no negative elements.

It is noteworthy that by choosing document-level granularity in our analysis, we assume that every review only carries a reviewer’s opinion on a single product (e.g., a movie or a TV show). Because when a document contains different people’s opinions on a single product or opinions of the reviewer on various products, the classification models can not correctly predict the general sentiment of the document. TextBlob is another excellent open-source library for performing NLP tasks with ease, including sentiment analysis. It also an a sentiment lexicon (in the form of an XML file) which it leverages to give both polarity and subjectivity scores. The subjectivity is a float within the range [0.0, 1.0] where 0.0 is very objective and 1.0 is very subjective.