Description:FIELD OF DISCLOSURE
[0001] The present disclosure relates generally relates to the field of artificial intelligence, natural language processing (NLP), and machine learning. More specifically, it pertains to a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter.
BACKGROUND OF THE DISCLOSURE
[0002] The rapid growth of social media platforms has transformed the way individuals and organizations communicate during times of crisis. Platforms such as Twitter have become primary channels for real-time dissemination of information, where users actively share opinions, experiences, and concerns during natural disasters, public health emergencies, political unrest, and other critical events. Unlike traditional media, where communication flows in a structured and delayed manner, Twitter provides instantaneous updates that reflect the emotional and cognitive states of people in affected communities. This dynamic flow of communication has created opportunities for crisis managers, emergency services, and governments to monitor public sentiment and respond effectively. However, the unstructured, noisy, and multilingual nature of Twitter data presents significant challenges for accurately detecting sentiments in real time.
[0003] Natural Language Processing (NLP) has historically played an important role in analyzing textual data. Sentiment analysis, a subdomain of NLP, has been applied to classify opinions and emotions expressed in social media posts. Early approaches in sentiment detection relied heavily on rule-based methods and manually constructed lexicons, where words and phrases were mapped to sentiment labels such as positive, negative, or neutral. While such methods provided foundational insights, they were often limited by their inability to capture context, sarcasm, idiomatic expressions, and emerging slang frequently used in online communication. Furthermore, lexicon-based systems struggled to adapt to domain-specific language shifts, particularly during crises, when new terminologies or hashtags emerge rapidly.
[0004] Machine learning introduced significant improvements to sentiment detection by leveraging supervised classification techniques. Algorithms such as Support Vector Machines (SVMs), Naïve Bayes, and logistic regression demonstrated greater flexibility in handling diverse datasets. These models, when trained on annotated corpora, were capable of learning statistical associations between textual features and sentiment classes. Feature engineering techniques, including n-grams, term frequency–inverse document frequency (TF-IDF), and part-of-speech tagging, were widely employed to improve classification accuracy. Despite their contributions, traditional machine learning models required extensive feature engineering, were sensitive to noise, and struggled with the contextual richness of short and ambiguous tweets.
[0005] The advent of deep learning further revolutionized sentiment detection by introducing models capable of learning hierarchical representations of text. Architectures such as Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs) enabled more effective modeling of sequential dependencies and contextual semantics. These models demonstrated superior performance over traditional machine learning by capturing subtleties in linguistic patterns. In crisis communication contexts, deep learning models provided more robust means of identifying urgency, fear, hope, and reassurance in social media discourse. However, the training of deep learning models required substantial computational resources and large labeled datasets, which are often difficult to obtain during unfolding crises.
[0006] The emergence of transformer-based models such as BERT, RoBERTa, GPT, and other large language models (LLMs) has marked another milestone in NLP. These models leverage self-attention mechanisms to process entire sequences in parallel, capturing both local and global contextual dependencies. Pre-trained on massive corpora, they can be fine-tuned for domain-specific tasks, significantly improving performance in sentiment analysis. In crisis communication, LLMs have shown promise in understanding nuanced expressions, handling code-switching between languages, and interpreting rapidly evolving terminology. Their ability to generalize across contexts has made them indispensable in real-time applications. Nevertheless, LLMs also introduce challenges such as high computational demands, latency in processing large-scale real-time streams, and concerns regarding interpretability of predictions.
[0007] Fusion-based approaches have emerged as an important direction to address these limitations. By combining multiple methods, such as lexicon-based features, traditional machine learning classifiers, and deep learning embeddings, researchers aim to exploit the complementary strengths of each paradigm. Fusion systems can integrate semantic, syntactic, and contextual cues, leading to more accurate and resilient sentiment detection. In the domain of crisis communication, where misinterpretation of public emotions can lead to inadequate or delayed responses, fusion strategies are particularly relevant. The dynamic nature of Twitter requires robust systems capable of real-time adaptability, handling massive data influx, and distinguishing between genuine and misleading information.
[0008] The increasing frequency of crises worldwide ranging from pandemics and natural disasters to sociopolitical movements has highlighted the necessity of advanced sentiment detection mechanisms. Organizations and authorities increasingly rely on public sentiment analysis to gauge societal mood, detect misinformation, and implement timely interventions. However, existing solutions often fall short when deployed in real-time, high-stakes environments. They may either be too slow to process the overwhelming volume of data or lack the accuracy required to capture subtle emotional shifts in communication.
[0009] Thus, in light of the above-stated discussion, there exists a need for a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter.
SUMMARY OF THE DISCLOSURE
[0010] The following is a summary description of illustrative embodiments of the invention. It is provided as a preface to assist those skilled in the art to more rapidly assimilate the detailed design discussion which ensues and is not intended in any way to limit the scope of the claims which are appended hereto in order to particularly point out the invention.
[0011] According to illustrative embodiments, the present disclosure focuses on a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter which overcomes the above-mentioned disadvantages or provide the users with a useful or commercial choice.
[0012] An objective of the present disclosure is to process high-volume, real-time Twitter streams efficiently, ensuring scalability and responsiveness of the sentiment detection system under rapidly changing crisis conditions.
[0013] Another objective of the present disclosure is to design and implement a fusion-based system that integrates traditional NLP techniques with advanced Large Language Models (LLMs) for accurate sentiment detection on Twitter during crisis communication.
[0014] Another objective of the present disclosure is to overcome the limitations of conventional NLP models in handling sarcasm, misinformation, evolving language, and nuanced expressions that frequently appear in social media crisis discourse.
[0015] Another objective of the present disclosure is to enable dynamic adaptation of the system so that it continuously learns and updates sentiment classification strategies in response to new linguistic patterns emerging during crises.
[0016] Another objective of the present disclosure is to provide granular sentiment insights (positive, negative, neutral, or fine-grained categories) that can guide policymakers, emergency responders, and organizations in improving crisis communication strategies.
[0017] Another objective of the present disclosure is to integrate real-time visualization and reporting mechanisms that allow stakeholders to monitor sentiment trends, misinformation propagation, and public emotional responses during unfolding crises.
[0018] Another objective of the present disclosure is to evaluate the effectiveness of the fusion-based approach by benchmarking the system against traditional NLP-only models in terms of accuracy, adaptability, and response speed.
[0019] Another objective of the present disclosure is to address challenges such as noisy data, short text context, and multilingual or code-mixed tweets commonly found on Twitter during crisis events.
[0020] Another objective of the present disclosure is to enhance trust and transparency in crisis communication by incorporating explainability features that clarify how the system interprets and classifies sentiment in ambiguous tweets.
[0021] Yet another objective of the present disclosure is to contribute to crisis management and public safety by providing decision-makers with real-time, data-driven sentiment insights that improve situational awareness and support timely interventions.
[0022] In light of the above, a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter comprises a data acquisition module configured to continuously collect Twitter posts and associated metadata in real time during crisis-related events. The system also includes a preprocessing module configured to clean, tokenize, and normalize the collected tweets. The system also includes a feature extraction module configured to extract linguistic, contextual, and semantic features from the preprocessed data using NLP techniques. The system also includes a large language model integration module configured to process the extracted features in association with pre-trained large language models. The system also includes a sentiment classification engine configured to assign sentiment labels. The system also includes a real-time adaptation unit configured to fine-tune sentiment predictions continuously in response to changing crisis-related discourse and emerging contextual signals. The system also includes a decision-support interface configured to provide authorities, emergency responders, and organizations with real-time insights on shifts in public sentiment, urgency levels, and emotional reactions, thereby enabling faster and empathetic crisis response.
[0023] In one embodiment, the data acquisition module is further configured to filter Twitter posts based on crisis-related keywords, hashtags, geolocation tags, and language preferences to ensure collection of relevant posts.
[0024] In one embodiment, the preprocessing module further comprises removing stop words, correcting spelling errors, and performing lemmatization and stemming to standardize the textual data.
[0025] In one embodiment, the feature extraction module employs natural language processing techniques selected from the group consisting of part-of-speech tagging, named entity recognition, sentiment lexicons, word embeddings, and contextual embeddings.
[0026] In one embodiment, the large language model integration module is configured to utilize one or more pre-trained models selected from the group consisting of GPT, BERT, RoBERTa, and T5 to process the extracted linguistic, contextual, and semantic features.
[0027] In one embodiment, the sentiment classification engine applies a supervised machine learning algorithm or a neural network-based classifier to assign sentiment labels including positive, negative, neutral, or mixed sentiments.
[0028] In one embodiment, the real-time adaptation unit continuously updates sentiment predictions by monitoring evolving crisis-related discourse, temporal patterns, and newly emerging hashtags, keywords, or contextual signals.
[0029] In one embodiment, the decision-support interface provides visualizations including trend graphs, heat maps, and dashboards to present real-time shifts in public sentiment, urgency levels, and emotional reactions to authorities, emergency responders, and organizations.
[0030] In one embodiment, the system is configured to generate automated alerts or notifications based on sudden shifts in public sentiment indicating critical urgency during a crisis event.
[0031] In one embodiment, the fusion of NLP and LLM-based insights is performed using an ensemble approach, attention mechanisms, or weighted scoring to improve accuracy and contextual relevance of sentiment detection.
[0032] These and other advantages will be apparent from the present application of the embodiments described herein.
[0033] The preceding is a simplified summary to provide an understanding of some embodiments of the present invention. This summary is neither an extensive nor exhaustive overview of the present invention and its various embodiments. The summary presents selected concepts of the embodiments of the present invention in a simplified form as an introduction to the more detailed description presented below. As will be appreciated, other embodiments of the present invention are possible utilizing, alone or in combination, one or more of the features set forth above or described in detail below.
[0034] These elements, together with the other aspects of the present disclosure and various features are pointed out with particularity in the claims annexed hereto and form a part of the present disclosure. For a better understanding of the present disclosure, its operating advantages, and the specified object attained by its uses, reference should be made to the accompanying drawings and descriptive matter in which there are illustrated exemplary embodiments of the present disclosure.
BRIEF DESCRIPTION OF THE DRAWINGS
[0035] To describe the technical solutions in the embodiments of the present disclosure or in the prior art more clearly, the following briefly describes the accompanying drawings required for describing the embodiments or the prior art. Apparently, the accompanying drawings in the following description merely show some embodiments of the present disclosure, and a person of ordinary skill in the art can derive other implementations from these accompanying drawings without creative efforts. All of the embodiments or the implementations shall fall within the protection scope of the present disclosure.
[0036] The advantages and features of the present disclosure will become better understood with reference to the following detailed description taken in conjunction with the accompanying drawing, in which:
[0037] FIG. 1 illustrates a flowchart outlining sequential step involved in a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter, in accordance with an exemplary embodiment of the present disclosure;
[0038] FIG. 2 illustrates the architectural flow diagram of a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter, in accordance with an exemplary embodiment of the present disclosure.
[0039] Like reference, numerals refer to like parts throughout the description of several views of the drawing;
[0040] The fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter, which like reference letters indicate corresponding parts in the various figures. It should be noted that the accompanying figure is intended to present illustrations of exemplary embodiments of the present disclosure. This figure is not intended to limit the scope of the present disclosure. It should also be noted that the accompanying figure is not necessarily drawn to scale.
DETAILED DESCRIPTION OF THE DISCLOSURE
[0041] The following is a detailed description of embodiments of the disclosure depicted in the accompanying drawings. The embodiments are in such detail as to communicate the disclosure. However, the amount of detail offered is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure.
[0042] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. It may be apparent to one skilled in the art that embodiments of the present disclosure may be practiced without some of these specific details.
[0043] Various terms as used herein are shown below. To the extent a term is used, it should be given the broadest definition persons in the pertinent art have given that term as reflected in printed publications and issued patents at the time of filing.
[0044] The terms “a” and “an” herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced items.
[0045] The terms “having”, “comprising”, “including”, and variations thereof signify the presence of a component.
[0046] Referring now to FIG. 1 to FIG. 2 to describe various exemplary embodiments of the present disclosure. FIG. 1 illustrates a flowchart outlining sequential step involved in a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter, in accordance with an exemplary embodiment of the present disclosure.
[0047] A fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter 100 comprises a data acquisition module 102 configured to continuously collect Twitter posts and associated metadata in real time during crisis-related events. The data acquisition module 102 is further configured to filter Twitter posts based on crisis-related keywords, hashtags, geolocation tags, and language preferences to ensure collection of relevant posts.
[0048] The system also includes a preprocessing module 104 configured to clean, tokenize, and normalize the collected tweets. The preprocessing module 104 further comprises removing stop words, correcting spelling errors, and performing lemmatization and stemming to standardize the textual data.
[0049] The system also includes a feature extraction module 106 configured to extract linguistic, contextual, and semantic features from the preprocessed data using NLP techniques. The feature extraction module 106 employs natural language processing techniques selected from the group consisting of part-of-speech tagging, named entity recognition, sentiment lexicons, word embeddings, and contextual embeddings. The fusion of NLP and LLM-based insights is performed using an ensemble approach, attention mechanisms, or weighted scoring to improve accuracy and contextual relevance of sentiment detection.
[0050] The system also includes a large language model integration module 108 configured to process the extracted features in association with pre-trained large language models. The large language model integration module 108 is configured to utilize one or more pre-trained models selected from the group consisting of GPT, BERT, RoBERTa, and T5 to process the extracted linguistic, contextual, and semantic features.
[0051] The system also includes a sentiment classification engine 110 configured to assign sentiment labels. The sentiment classification engine 110 applies a supervised machine learning algorithm or a neural network-based classifier to assign sentiment labels including positive, negative, neutral, or mixed sentiments.
[0052] The system also includes a real-time adaptation unit 112 configured to fine-tune sentiment predictions continuously in response to changing crisis-related discourse and emerging contextual signals. The real-time adaptation unit 112 continuously updates sentiment predictions by monitoring evolving crisis-related discourse, temporal patterns, and newly emerging hashtags, keywords, or contextual signals.
[0053] The system also includes a decision-support interface 114 configured to provide authorities, emergency responders, and organizations with real-time insights on shifts in public sentiment, urgency levels, and emotional reactions, thereby enabling faster and empathetic crisis response. The decision-support interface 114 provides visualizations including trend graphs, heat maps, and dashboards to present real-time shifts in public sentiment, urgency levels, and emotional reactions to authorities, emergency responders, and organizations.
[0054] The system is configured to generate automated alerts or notifications based on sudden shifts in public sentiment indicating critical urgency during a crisis event.
[0055] FIG. 1 illustrates a flowchart outlining sequential step involved in a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter.
[0056] At 102, the system begins with the data acquisition module, which is continuously engaged in collecting Twitter posts along with their associated metadata in real time. This ensures that the system remains current with the rapidly evolving information landscape that characterizes crises, capturing tweets related to emergencies, natural disasters, public health incidents, or social unrest, and assembling the necessary context to inform subsequent analytical processes. The uninterrupted collection of data is critical, as it allows the system to maintain a comprehensive and up-to-date repository of public communications that can be leveraged for sentiment evaluation.
[0057] At 104, once the data is acquired, it flows into the preprocessing module, where it undergoes a series of essential cleaning and normalization operations. This module systematically removes noise, such as irrelevant characters, hyperlinks, or non-standard symbols, while standardizing the textual content through tokenization and normalization. These preprocessing steps ensure that the raw tweets are transformed into a consistent format suitable for analysis, reducing potential inaccuracies caused by informal language, abbreviations, or inconsistent structures common in social media posts. The cleaned data serves as a robust foundation for the extraction of meaningful features, enabling more precise and reliable sentiment analysis.
[0058] At 106, the feature extraction module then processes the preprocessed tweets to derive linguistic, contextual, and semantic attributes that reflect both the content and the underlying intent of the messages. Using advanced NLP techniques, this module captures syntactic structures, word embeddings, sentiment-bearing phrases, topic associations, and contextual cues that indicate emotional tone or urgency. By distilling the rich textual data into a structured set of features, the system creates a representation that can be effectively interpreted by downstream analytical engines, bridging the gap between raw social media text and computational sentiment understanding.
[0059] At 108, the extracted features are fed into the large language model integration module, which leverages pre-trained large language models to interpret complex patterns, contextual dependencies, and nuanced expressions within the tweets. This module enables the system to account for subtleties in language, including sarcasm, idiomatic expressions, and culturally specific references, which are often challenging for conventional models to capture. The synergy between NLP-derived features and LLM processing enhances the depth and accuracy of sentiment recognition, ensuring that even sophisticated linguistic cues are effectively incorporated into the analysis.
[0060] At 110, the sentiment classification engine is responsible for assigning precise sentiment labels to each tweet. By utilizing both the outputs from the LLM integration and the structured features obtained during extraction, this engine categorizes messages according to predefined sentiment classes, such as positive, negative, or neutral, while also recognizing gradations of intensity or emotional valence. This classification process transforms the nuanced understanding generated by the language models into actionable insights that can guide real-world responses.
[0061] At 112, recognizing that crisis-related discourse evolves dynamically, the system incorporates a real-time adaptation unit that continuously fine-tunes sentiment predictions in response to emerging trends and changing conversational patterns. This unit monitors shifts in language use, sudden spikes in discussion, and contextual signals, adjusting the model’s parameters to maintain accuracy and relevance. By adapting in real time, the system ensures that authorities and responders receive sentiment assessments that reflect the current state of public discourse, rather than static or outdated interpretations.
[0062] At 114, the decision-support interface serves as the point of interaction for stakeholders, providing authorities, emergency responders, and organizations with actionable insights derived from the sentiment analysis. This interface presents comprehensive visualizations, trend indicators, and urgency metrics, allowing decision-makers to quickly understand shifts in public emotion, identify areas of concern, and respond in an informed, empathetic, and timely manner. By bridging the gap between sophisticated computational analysis and practical crisis management, the system facilitates proactive and responsive strategies that can mitigate the impact of emergencies and enhance public communication efficacy.
[0063] FIG. 2 illustrates the architectural flow diagram of a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter.
[0064] The process begins with the ingestion of live tweet data, where information from the Twitter platform is collected in real time. This step is crucial as it forms the raw input for the system, capturing unstructured text data that users generate continuously. The ingested data may include a wide range of linguistic styles, abbreviations, and even noisy elements such as emojis or hashtags, which makes subsequent processing essential.
[0065] After data ingestion, the next stage involves Natural Language Processing (NLP). Within this layer, multiple computational techniques are applied to transform raw textual tweets into meaningful representations. Feature extraction identifies key aspects of the text, such as keywords, sentence structures, and semantic relationships. Emotion detection focuses on identifying underlying human emotions such as joy, anger, sadness, or fear hidden in the language of tweets. Furthermore, language pattern analysis is applied to uncover syntactic and semantic structures, recognizing trends such as sarcasm, context-dependent meanings, or frequent co-occurrence of words. This NLP stage acts as a bridge between unstructured text and structured, analyzable data.
[0066] The refined data is then passed into a contextual refinement stage powered by a large language model (LLM). Here, the LLM ensures that the detected features and emotions are interpreted within the proper context. For instance, words that might otherwise appear negative in isolation could be reclassified as neutral or even positive when considered in the broader sentence structure or conversational context. The outcome of this refinement is a sentiment classification of tweets into three categories: positive, negative, and neutral. These categories provide a simplified but powerful representation of public opinion, enabling further applications such as trend analysis, brand monitoring, or policy feedback evaluation.
[0067] While the invention has been described in connection with what is presently considered to be the most practical and various embodiments, it will be understood that the invention is not to be limited to the disclosed embodiments, but on the contrary, is intended to cover various modifications and equivalent arrangements included within the scope of the appended claims.
[0068] A person of ordinary skill in the art may be aware that, in combination with the examples described in the embodiments disclosed in this specification, units and algorithm steps may be implemented by electronic hardware, computer software, or a combination thereof.
[0069] The foregoing descriptions of specific embodiments of the present disclosure have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed, and many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described to best explain the principles of the present disclosure and its practical application, and to thereby enable others skilled in the art to best utilize the present disclosure and various embodiments with various modifications as are suited to the particular use contemplated. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but such omissions and substitutions are intended to cover the application or implementation without departing from the scope of the present disclosure.
[0070] Disjunctive language such as the phrase “at least one of X, Y, Z,” unless specifically stated otherwise, is otherwise understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and/or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.
[0071] In a case that no conflict occurs, the embodiments in the present disclosure and the features in the embodiments may be mutually combined. The foregoing descriptions are merely specific implementations of the present disclosure, but are not intended to limit the protection scope of the present disclosure. Any variation or replacement readily figured out by a person skilled in the art within the technical scope disclosed in the present disclosure shall fall within the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims
, Claims:Disclosed herein is a fusion-based NLP and LLMs system for real-time sentiment detection in crisis communication on twitter (100) comprises a data acquisition module (102) configured to continuously collect Twitter posts and associated metadata. The system also includes a preprocessing module (104) configured to clean, tokenize, and normalize the collected tweets. The system also includes a feature extraction module (106) configured to extract linguistic, contextual, and semantic features from the preprocessed data. The system also includes a large language model integration module (108) configured to process the extracted features in association with pre-trained large language models. The system also includes a sentiment classification engine (110) configured to assign sentiment labels. The system also includes a real-time adaptation unit (112) configured to fine-tune sentiment predictions. The system also includes a decision-support interface (114) configured to provide authorities, emergency responders, and organizations with real-time insights.