Description:FIELD OF THE INVENTION
The present invention relates generally to the field of natural language processing (NLP) and artificial intelligence (AI). More particularly, the invention is directed to a system and method for sentiment and emotion analysis of online texts using transfer learning techniques in large-scale pre-trained transformer-based models. The invention enables multi-label classification of emotions, sentiment polarity determination, and generation of interpretive explanations from user-generated online content such as social media posts, product reviews, blogs, and forums.
BACKGROUND OF THE INVENTION
With more and more people sharing their thoughts and feelings online through social media, forums, and review sites, we now have a huge amount of text that reflects different emotions and opinions. This kind of information can be really useful for understanding public mood, mental health trends, and even customer satisfaction. But figuring out exactly what emotions or sentiments are being expressed in all this text is tricky, especially because online language can be casual, sarcastic, or very dependent on context.
Older methods for analyzing sentiment often use fixed rules or basic machine learning, but they struggle with understanding deeper meanings or emotional tone. That is where new AI models like T5 and GPT-4 come in—they are trained on massive amounts of data and can understand language in a much more sophisticated way.
This project focuses on using these advanced models to improve how we detect emotions (like happiness, anger, or sadness) and overall sentiment (positive, negative, or neutral) in online text. By training them on special datasets labeled with emotions, we want to see how well they can do compared to older techniques.
The challenge is not just to get accurate results, but to make sure the models work well across different kinds of online content. We also have to deal with issues like unbalanced data, understanding context, and making sure the models run efficiently.
US20220405485A1: A text-based real-time communication interface, such as a chatbot, is presented to a user for the exchange of customer support information. A user's freeform text input is analyzed using machine learning algorithms to derive the meaning of the input text as well as to determine the user sentiment expressed therein. These determinations may be further supported by signals extracted from session-based activity, which signals can be used to infer the intended workflow of the user and whether or not that workflow was achieved. The expressed user sentiment is considered along with other historical or session-based user data to generate tailored questions and responses to be delivered in real-time to the user. The responses are displayed to the user along with information that routes the user to a workflow resolution.
US20190171660: A system and method for improved categorization and sentiment analysis which is fed textual data such as transcriptions or collated data from a network enabled service, or some other source, which then segments textual data into chunks, parses the data chunks, and analyzes it using a plurality of techniques and metadata gathering methods to determine the sentiment of participating individuals concerning entities mentioned in the textual data and to categorize the discussions, for the purpose of taking actions to improve business outcomes.
The present invention addresses the limitations of existing sentiment and emotion analysis systems applied to online texts. Current solutions either provide only coarse sentiment polarity (positive, negative, neutral) or limited emotion sets that lack adaptability to different domains. They fail to capture subtle emotions, sarcasm, irony, and context-heavy meanings common in social media, reviews, and forums. Many existing systems rely solely on classification without explanation, functioning as black-box models, which reduces trust and applicability in sensitive domains like healthcare, legal analysis, and journalism. The invention solves these problems by leveraging transfer learning in advanced transformer architectures, specifically T5 and GPT-4, combined with prompt engineering, to perform multi-task emotion and sentiment analysis. The system not only classifies sentiment and emotions with higher accuracy, but also generates explainable rationales, summaries, and context-aware outputs, thereby ensuring interpretability, adaptability, and real-world applicability.
SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
At the top, the diagram begins with the title of the proposed invention, establishing the focus on using advanced transformer models (T5 and GPT-4) for analyzing emotional and sentiment content in online texts.
BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
FIGURE 1: SYSTEM ARCHITECTURE
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein 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 scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a",” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may, in fact, be executed concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In addition, the descriptions of "first", "second", “third”, and the like in the present invention are used for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, features defining "first" and "second" may include at least one of the features, either explicitly or implicitly.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The invention relates to a system and method for advanced emotion and sentiment analysis of online texts using transfer learning techniques applied to transformer models T5 and GPT-4.
Online texts include user-generated content from social media, product reviews, blogs, and discussion forums, which often contain sarcasm, irony, abbreviations, and context-heavy expressions.
The invention’s input module captures such online texts in raw form, requiring minimal preprocessing, since transformer models can process unstructured text effectively.
A pre-trained T5 model is employed for classification-oriented tasks. It reformulates emotion and sentiment recognition into a text-to-text problem, thereby allowing flexible labeling and fine-grained output generation.
A GPT-4 model is employed for generative tasks, including explanation generation, contextual reasoning, and sentiment-based summarization of online text.
The hybrid framework combines the classification power of T5 with the generative reasoning power of GPT-4, ensuring robust detection of subtle emotions and better handling of sarcasm, idioms, and complex narratives.
Transfer learning enables the system to adapt to different domains with minimal fine-tuning, using small, domain-specific labeled datasets.
The system architecture includes a training and fine-tuning module, which applies supervised learning on emotion-annotated datasets to refine model performance.
The invention further includes a prompt engineering module, where specialized prompts guide GPT-4 and T5 models to extract targeted emotions and sentiments more effectively.
The system supports multi-label classification, where texts can be labeled with multiple emotions (e.g., “angry but hopeful”) rather than single polarity classification.
The invention’s explanation generation unit provides interpretability by generating natural language justifications, such as highlighting keywords, context, or semantic cues leading to classification.
The invention enables summarization of sentiment trends, producing concise reports from large volumes of text to support decision-making.
Outputs are delivered in three formats: emotion classification (specific emotions), sentiment labels (positive, negative, neutral), and textual explanations.
Unlike conventional solutions, the system supports dynamic emotional taxonomies, allowing users to define or expand emotion categories based on application domains.
The invention is scalable, capable of processing large volumes of online data across multiple languages due to the multilingual training capabilities of T5 and GPT-4.
The deployment configuration can be cloud-based or edge-based depending on application requirements, making it suitable for both large-scale analytics and localized decision support.
The system includes an evaluation and feedback loop, which continuously improves classification performance using user feedback and active learning.
Security and privacy are maintained by anonymizing user text data and applying encryption for sensitive domains such as healthcare and finance.
The invention improves efficiency by reducing the need for extensive labeled data, instead leveraging few-shot or zero-shot learning capabilities of GPT-4.
Overall, the invention achieves higher accuracy, better interpretability, and stronger adaptability compared to existing emotion and sentiment analysis systems.
The invention discloses a hybrid system and method for sentiment and emotion analysis of online texts using transfer learning in advanced transformer-based models, specifically T5 and GPT-4. The invention integrates both generative and classification capabilities, allowing the models to detect emotions, classify sentiments, and produce natural language explanations.
At the core, the system ingests online text data from various sources such as social media, product reviews, and discussion forums. Pre-trained T5 and GPT-4 models are fine-tuned or guided with prompt engineering to detect nuanced emotions such as joy, sadness, anger, fear, or anxiety, in addition to conventional sentiment categories. Unlike conventional methods, the invention unifies multiple tasks—classification, summarization, and explanation—within a single framework.
The invention introduces domain adaptability by leveraging transfer learning, enabling accurate performance across multiple domains such as healthcare, finance, and customer service with minimal fine-tuning. Additionally, the invention provides explainable AI outputs, where the system not only gives emotion or sentiment labels but also generates textual reasoning. This enhances trustworthiness, reduces misclassification, and supports automated decision-making in professional applications.
At the top, the diagram begins with the title of the proposed invention, establishing the focus on using advanced transformer models (T5 and GPT-4) for analyzing emotional and sentiment content in online texts.
1. Input Stage: The next layer shows the two primary inputs:
o Online Texts: User-generated content from platforms like social media, forums, or reviews.
o GPT-4 Model: A pre-trained transformer model capable of understanding and generating human-like language.
2. Processing Stage: These inputs are funnelled into a central process block labeled "Emotion and Sentiment Analysis." This represents the core activity where the transformer models analyse the text to extract emotional tone and sentiment.
3. Output Stage: The analysis leads to three key outcomes:
o Summaries: Condensed versions of the analysed text highlighting key emotional and sentiment insights.
o Emotion Classification: Identification and labelling of specific emotions expressed in the text (e.g., joy, anger, sadness).
o Sentiment Explanation: A rationale or explanation of why the model classified a piece of text with a certain sentiment, adding interpretability to the results.
Key Goals
i.To improve the precision of user-generated text sentiment and emotion identification.
ii.To employ a generative-classification hybrid model to capture complex and context-driven emotions.
iii.To combine contextual generation (from GPT-4) with interpretability (from T5) in order to get beyond the drawbacks of current models.
iv. To enhance domain generalization through minimal supervision and transfer learning.
1.Online Texts -This is the raw input collected from various online sources such as social media posts, product reviews or comments, texts often contain emotional expressions.
2. Pre-trained T5/GPT-4 Model - The core of the system uses pre-trained transformer models (T5 or GPT-4) that understand language very well due to extensive training on large -text corpora. and form the foundation which are ready to be adapted to the specific task.
3.Fine-tuning- The general-purpose models are fine-tuned on specific emotion- or sentiment-labeled datasets.
4.Sentiment & Emotion Analysis- This is the main analysis step where the fine-tuned model processes input texts. It extracts emotional content and sentiment.
5. Outputs:i. Classification-Identifies specific emotions like happy, anger, sadness.
ii. Sentiment Labels-Assigns general sentiment categories likepositive, negative, or neutral.
iii. Explanation-The model can also provide a natural language explanation of why a particular emotion or sentiment was assigned.
Understanding human feelings and sentiments is made more difficultby the rapid growth of online content. The diversity, and shades of online content can make traditional sentiment and emotion research methods are not good or advanced enough. The T5 and GPT-4 advanced generative transformer models are used in this study to overcome those defaults.
Best Method of Working
The best method of working the invention involves configuring the system with both T5 and GPT-4 pre-trained models. Online text is collected via the input module and preprocessed minimally. The T5 model is fine-tuned with labeled emotion datasets to handle classification tasks, while GPT-4 is prompted to perform contextual explanation and reasoning tasks. During operation, a user provides text input, which is simultaneously analyzed by both models. The T5 model outputs emotion/sentiment labels, while GPT-4 generates natural language explanations and concise summaries. A feedback loop refines the models over time. This hybrid setup ensures accurate classification, interpretability, and domain adaptability with minimal additional training.
, Claims:1. A system for emotion and sentiment analysis of online texts, comprising:
an input module for receiving online text data;
a pre-trained T5 transformer model configured for sentiment and emotion classification;
a GPT-4 model configured for generative reasoning and explanation;
a prompt engineering module for guiding classification and generative tasks;
a hybrid framework integrating classification and generative outputs; and
an output module for generating emotion labels, sentiment labels, and natural language explanations.
2. The system as claimed in claim 1, wherein the input module processes raw online texts including social media posts, reviews, and forum discussions with minimal preprocessing.
3. The system as claimed in claim 1, wherein the hybrid framework performs multi-label classification for multiple simultaneous emotions.
4. The system as claimed in claim 1, wherein the GPT-4 model generates natural language explanations to provide interpretability of sentiment classification.
5. The system as claimed in claim 1, wherein the T5 model reformulates sentiment and emotion classification as a text-to-text transformation problem.
6. The system as claimed in claim 1, wherein the prompt engineering module enables domain-specific adaptation for healthcare, finance, or customer service.
7. The system as claimed in claim 1, wherein the system performs summarization of sentiment trends across large datasets.
8. The system as claimed in claim 1, wherein the output module supports dynamic emotional taxonomies customizable by end-users.
9. The system as claimed in claim 1, wherein the hybrid framework is trained using transfer learning with minimal labeled datasets.
10. A method for emotion and sentiment analysis of online texts using transfer learning, comprising the steps of:
receiving online text data;
processing the data using a pre-trained T5 transformer model for emotion and sentiment classification;
processing the data using a GPT-4 model for generating explanations and summaries;
integrating classification and generative outputs through a hybrid framework; and
outputting sentiment labels, emotion labels, and interpretive explanations.