Amharic political sentiment analysis using deep learning approaches Scientific Reports

This hypothesis has not been fully supported, since the four sub-corpora have proved to be similarly intense in the high levels of emotional activity recorded, thus our initial assumption that economic reports are highly charged in emotional terms is not confirmed. Two researchers attempted to design a deep learning model for Amharic sentiment analysis. The CNN model designed by Alemu and Getachew8 was overfitted and did not generalize well from training data to unseen data. This problem was solved in this research by adjusting the hyperparameter of the model and shift the model from overfitted to fit that can generalize well to unseen data. The CNN-Bi-LSTM model designed in this study outperforms the work of Fikre19 LSTM model with a 5% increase in performance. This work has a major contribution to update the state-of-the-art Amharic sentiment analysis with improved performance.

From the data visualization, we observed that the YouTube users had an opinion for the conflicted party to solve it peacefully. In this section, we also understand that so many users use YouTube to express their opinions related to wars. This shows that any conflicted country should view YouTube users for their decision. To categorize YouTube users’ opinions, we developed deep learning models, which include LSTM, GRU, Bi-LSTM, and Hybrid (CNN-Bi-LSTM).

Unveiling the nature of interaction between semantics and phonology in lexical access based on multilayer networks

However, these metrics might be indicating that the model is predicting more articles as positive. We can see that the spread of sentiment polarity is much higher in sports and world as compared to technology where a lot of the articles seem to be having a negative polarity. This is not an exhaustive list of lexicons that can be leveraged for sentiment analysis, and there are several other lexicons which can be easily obtained from the Internet. From the preceding output, you can see that our data points are sentences that are already annotated with phrases and POS tags metadata that will be useful in training our shallow parser model.

Term Frequency-Inverse Document Frequency (TF-IDF) is a weighting schema that uses term frequency and inverse document frequency to discriminate items29. As previously said, the Urdu language has a morphological structure that is highly unique, exceedingly rich, and complex when compared to other resource-rich languages. Urdu is a blend of several languages, including Hindi, Arabic, Turkish, Persian, and Sanskrit, and contains loan words from these languages. Other reasons for incorrect classifications include the fact that the normalization of Urdu text is not yet perfect. To tokenize Urdu text, spaces between words must be removed/inserted because the boundary between words is not visibly apparent.

For example, the frequencies of agents (A0) and discourse markers (DIS) in CT are higher than those in both ES and CO, suggesting that the explicitation in these two roles is both S-oriented and T-oriented. In other words, there is an additional force that drives the translated language away from both the source and target language systems, and this force could be pivotal in shaping translated language as “the third language” or “the third code”. For the exploration of S-universals, ES are compared with CT in Yiyan English-Chinese Parallel Corpus (Yiyan Corpus) (Xu & Xu, 2021). Yiyan Corpus is a million-word balanced English-Chinese parallel corpus created according to the standard of the Brown Corpus.

Urdu datasets and machine learning techniques

Between 1966 and 1976, after a decade of the Cultural Revolution, the Chinese government recognized the importance of stability for the country’s economic development. In 1989, one of Deng Xiaoping’s basic tenets was “Stability is of paramount importance” (稳定压倒一切, wen ding ya dao yi qie) (Deng, 1994). Consequently, “stability” has become one of China’s most frequently used political keywords. Looking at SBS components, we can notice that all of them are equally accurate in forecasting Personal Climate, while connectivity is the best performer also for Economic and Current Climate, for this second variable together with diversity. Notice that both AR and BERT models are always statistically different with respect to the best performer, while AR(2) + Sentiment performs worse than the best model for 3 variables out of 5. Table 4 illustrates the mean square forecasting errors (MSFEs) relative to the AR(2) forecasts.

• Because of increasing interest in SA, businesses are interested in driving campaigns, having more clients, overcoming their weaknesses, and winning marketing tactics.
• Meltwater features intuitive dashboards, customizable searches, and visualizations.
• Both proposed models, leveraging LibreTranslate and Google Translate respectively, exhibit better accuracy and precision, surpassing 84% and 80%, respectively.
• In our review, we report the latest research trends, cover different data sources and illness types, and summarize existing machine learning methods and deep learning methods used on this task.
• Businesses can use machine-learning-based sentiment analysis software to examine this speech and text for positive or negative sentiment about the brand.

EWeek stays on the cutting edge of technology news and IT trends through interviews and expert analysis. Gain insight from top innovators and thought leaders in the fields of IT, business, enterprise software, startups, and more. NLTK is great for educators and researchers because it provides a broad range of NLP tools and access to a variety of text corpora.

The classification of sentiment analysis includes several states like positive, negative, Mixed Feelings and unknown state. Similarly for offensive language identification the states include not-offensive, offensive untargeted, offensive targeted insult group, offensive targeted insult individual and offensive targeted insult other. Finally, the results are classified into respective states and the models are evaluated using performance metrics like precision, recall, accuracy and f1 score. Sentiment analysis is a Natural Language Processing (NLP) task concerned with opinions, attitudes, emotions, and feelings.

However, the performance we obtained was worse than the non-recurrent version we reported in the result section. This is probably due to the limited number of training samples, which are insufficient to optimize the more complex recurrent model. To nowcast CCI indexes, we trained a neural network that took the BERT encoding of the current week and the last available CCI index score (of the previous month) as input. The network comprised a hidden layer with ReLU activation, a dropout layer for regularization, and an output layer with linear activation that predicts the CCI index. You can foun additiona information about ai customer service and artificial intelligence and NLP. From the Consumer Confidence Climate survey, we extracted economic keywords that were recurring in the survey’s questions. We then extended this list by adding other relevant keywords that matched the economic literature and the independent assessment of three economics experts.

Corpus generation

Through the application of quantitative methods and computational power, these studies aim to uncover insights regarding the structure, trends, and patterns within the literature. The field of digital humanities offers diverse and substantial perspectives on social situations. While it is important to note that predictions made in this field may not be applicable to the entire world, they hold significance for specific research objects. For example, in computational linguistics research, the lexicons used in emotion analysis are closely linked to relevant concepts and provide accurate results for interpreting context. However, it is important to acknowledge that embedded dictionaries and biases may introduce exceptions that cannot be completely avoided. Nonetheless, computational literary studies offer advantages such as quick interpretation, analysis, and prediction on extensive datasets (Kim and Klinger, 2018).

Furthermore, Sawhney et al. introduced the PHASE model166, which learns the chronological emotional progression of a user by a new time-sensitive emotion LSTM and also Hyperbolic Graph Convolution Networks167. It also learns the chronological emotional spectrum of a user by using BERT fine-tuned for emotions as well as a heterogeneous social network graph. As mentioned above, machine learning-based models rely heavily on feature engineering and feature extraction.

For Arabic, the recall scores are notably high across various combinations, indicating effective sentiment analysis for this language. These findings suggest that the proposed ensemble model, along with GPT-3, holds promise for improving recall in multilingual sentiment analysis tasks across ChatGPT App diverse linguistic contexts. The work in11, systematically investigates the translation to English and analyzes the translated text for sentiment within the context of sentiment analysis. Arabic social media posts were employed as representative examples of the focus language text.

To do so, we built an LDA model to extract feature vectors from each day’s news and then deployed logistic regression to predict the direction of market volatility the next day. To measure our classifier performance, we used the standard measures of accuracy, recall, precision, and F1 score. All these measures were obtained using the well-known Python Scikit-learn module4. Our causality testing exhibited no reliable causality between the sentiment scores and the FTSE100 return with any lags. We found that causality slightly increased at a time lag of 2 days but it remained statistically insignificant. Vice versa Granger’s text found statistical significance in negative returns causing negative sentiment, as expected.

Transformers have become the backbone of various state-of-the-art models in NLP, including BERT, GPT and T5 (Text-to-Text Transfer Transformer), among others. They excel in tasks such as language modeling, machine translation, text generation and question answering. The success of Word2Vec and GloVe have inspired further research into more sophisticated language representation models, such as FastText, BERT and GPT.

Finally, a long short-term memory-gated recurrent unit (LSTM-GRU) deep learning model is built to classify the sentiment characteristics that induce sexual harassment. The proposed model achieved an accuracy of 75.8% while outperforming five other algorithms. Additionally, a sentiment classification with three labels—negative, positive, and neutral—was developed using an LSTM-GRU RNN deep learning model. Most statements, even those involving physical sexual harassment, which had greater levels of sexual harassment, had negative sentiments, according to lexicon-based sentiment analysis. This study contributes to the field of text mining by providing a novel approach to identifying instances of sexual harassment in literature in English from the Middle East.

• In other words, it will keep the points of majority class that’s most different to the minority class.
• They obtained a 56% accuracy in predicting directional stock market volatility on the arrival of new information.
• 1, extremely long roles can be attributed to multiple substructures nested within the semantic role, such as A1 in Structure 1 (Fig. 1) in the English sentence, which contains three sub-structures.
• If you want to know more about Tf-Idf, and how it extracts features from text, you can check my old post, “Another Twitter Sentiment Analysis with Python-Part5”.

Only 650 movie reviews are included in the C1 dataset, with each review averaging 264 words in length. The other dataset named C2, contains 700 reviews about refrigerators, air conditions, and televisions. IBM Watson NLU stands out in ChatGPT terms of flexibility and customization within a larger data ecosystem. Users can extract data from large volumes of unstructured data, and its built-in sentiment analysis tools can be used to analyze nuances within industry jargon.

This enables developers and businesses to continuously improve their NLP models’ performance through sequences of reward-based training iterations. Such learning models thus improve NLP-based applications such as healthcare and translation software, chatbots, and more. German startup deepset develops a cloud-based software-as-a-service (SaaS) platform for NLP applications. It features all the core components semantic analysis of text necessary to build, compose, and deploy custom natural language interfaces, pipelines, and services. The startup’s NLP framework, Haystack, combines transformer-based language models and a pipeline-oriented structure to create scalable semantic search systems. Moreover, the quick iteration, evaluation, and model comparison features reduce the cost for companies to build natural language products.

Tokenization is the process of separating raw data into sentence or word segments, each of which is referred to as a token. In this study, we employed the Natural Language Toolkit (NLTK) package to tokenize words. Tokenization is followed by lowering the casing, which is the process of turning each letter in the data into lowercase. This phase prevents the same word from being vectorized in several forms due to differences in writing styles. The first layer in a neural network is the input layer, which receives information, data, signals, or features from the outside world. Cdiscount, an online retailer of goods and services, uses semantic analysis to analyze and understand online customer reviews.

People convey different emotions to give responses and reactions according to different circumstances. Emotion detection has been proven to be beneficial in identifying criminal motivations and psychosocial interventions (Guo, 2022). Sentiment and emotions can be classified based on the domain knowledge and context using NLP techniques, including statistics, machine learning and deep learning approaches. The results presented in this study provide strong evidence that foreign language sentiments can be analyzed by translating them into English, which serves as the base language.

The language conveys a clear or implicit hint that the speaker is depressed, angry, nervous, or violent in some way is presented in negative class labels. Mixed-Feelings are indicated by perceiving both positive and negative emotions, either explicitly or implicitly. Finally, an unknown state label is used to denote the text that is unable to predict either as positive or negative25. We illustrate the efficacy of GML by the examples from CR as shown in Table 5 and Figure 7. On \(t_1\), both GML and the deep learning model give the correct label; however, on all the other examples, GML gives the correct labels while the deep learning model mispredicts. In Figure 7, the four subfigures show the constructed factor subgraphs of the examples respectively.

Additionally, novel end-to-end methods for pairing aspect and opinion terms have moved beyond sequence tagging to refine ABSA further. These strides are streamlining sentiment analysis and deepening our comprehension of sentiment expression in text55,56,57,58,59. To effectively navigate the complex landscape of ABSA, the field has increasingly relied on the advanced capabilities of deep learning. Neural sequential models like Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU) have set the stage by adeptly capturing the semantics of textual reviews36,37,38. These models contextualize the sequence of words, identifying the sentiment-bearing elements within. The Transformer architecture, with its innovative self-attention mechanisms, along with Embeddings from Language Models (ELMo), has further refined the semantic interpretation of texts39,40,41.

The proposed solution leverages the existing DNN models to extract polarity-aware binary relation features, which are then used to enable effective gradual knowledge conveyance. Our extensive experiments on the benchmark datasets have shown that it achieves the state-of-the-art performance. Our work clearly demonstrates that gradual machine learning, in collaboration with DNN for feature extraction, can perform better than pure deep learning solutions on sentence-level sentiment analysis. 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.

If we have enough examples, we can even train a deep learning model for better performance. 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).

The second-best performance was obtained by combining LDA2Vec embedding and implicit incongruity features. The bag of Word (BOW) approach constructs a vector representation of a document based on the term frequency. However, a drawback of BOW representation is that word order is not preserved, resulting in losing the semantic associations between words. The representation vectors are sparse, with too many dimensions equal to the corpus vocabulary size31. Homonymy means the existence of two or more words with the same spelling or pronunciation but different meanings and origins.

The Maslow’s hierarchy of needs theory is applied to guide the consistent sentiment annotation. The domain lexicon is integrated into the feature fusion layer of the RoBERTa-FF-BiLSTM model to fully learn the semantic features of word information, character information, and context information of danmaku texts and perform sentiment classification. The limitations of this paper are that the construction of the domain lexicon still requires manual participation and review, the semantic information of danmaku video content and the positive case preference are ignored. Furthermore, the size of available annotated datasets is insufficient for successful sentiment analysis. However, the majority of the datasets and reviews from limited domains are only from negative and positive classes. To address this issue, this work focuses on the creation of an Urdu text corpus that includes sentences from several genres.