Context-Aware AI Chatbot Using Transformer-Based Models for

Intelligent User Interactions

Pooja S, Gokul G, Linkesh Mani K, Raj Kumar A S and Amutha Bharathi R

Department of AIDS, Karpagam Academy of Higher Education, Coimbatore, India

Keywords: EfficientNet, Vision Transformers (ViT), Red Piranha Optimization (RPO).

Abstract: Chatbots have transformed user interaction with technology by offering immediate, automated support across

multiple sectors. Advancements in artificial intelligence (AI) and natural language processing (NLP) are

enhancing the efficiency and contextual awareness of contemporary chatbots. This study introduces an AI-

driven chatbot system utilizing machine learning and natural language processing to facilitate effective and

user-friendly conversations. The suggested chatbot utilizes sophisticated methodologies, including

Transformer-based models like BERT for intent recognition and Named Entity Recognition (NER), as well

as GPT for producing dynamic, human-like responses. The system is engineered to manage a variety of

conversational duties, ranging from addressing frequently asked questions to executing transactional

transactions. It incorporates preprocessing methods such as tokenization and text normalization to improve

input comprehension and employs embeddings for contextual understanding. The chatbot employs

conversational state tracking with recurrent neural networks (RNNs) or memory-augmented Transformers to

facilitate coherent and contextually aware multi-turn discussions. Response generation employs a hybrid

methodology, integrating template-driven answers for structured inquiries with dynamic replies for open-

ended questions. The chatbot backend interfaces with REST APIs to retrieve external data, guaranteeing real-

time capabilities for user-specific operations such as reservations or database inquiries. The system is

implemented on scalable cloud platforms and is available through several channels, including online and

mobile applications, as well as messaging networks such as WhatsApp and Telegram. Despite its efficacy,

constraints encompass computational complexity for real-time Transformer-based inference and reliance on

high-quality training data. Future improvements involve utilizing hybrid deep learning models to enhance

scalability and robustness.

1 INTRODUCTION

Chatbots, commonly referred to as chatterbots, are

software agents that emulate human conversation

through text or voice messaging. One of the primary

objectives of Chatbot has consistently been to

emulate an intelligent human, thereby obscuring its

true nature from others. The proliferation of many

chatbot architectures and functionalities has

significantly broadened their application. These

conversational agents can deceive consumers into

believing they are interacting with a human, although

they are significantly constrained in enhancing their

knowledge base in real-time. The chatbot employs

artificial intelligence and deep learning techniques to

comprehend user input and generate a relevant

answer. Furthermore, they engage with humans

through natural language, utilizing many applications

of chatbots, including medical chatbots and contact

centers. A chatbot could assist physicians, nurses,

patients, or their relatives. Enhanced organization of

patient data, medication oversight, assistance during

emergencies or with first aid, and provision of

solutions for minor medical concerns: these scenarios

exemplify the potential for chatbots to alleviate the

workload of healthcare workers.

Chatbots have revolutionized digital

communication by facilitating effortless interactions

between individuals and systems. AI-driven solutions

are extensively employed in sectors such customer

service, healthcare, and education to deliver

immediate, dependable assistance. With the

emergence of sophisticated natural language

S, P., G, G., Mani K, L., A S, R. K. and Bharathi R, A.

Context-Aware AI Chatbot Using Transformer-Based Models for Intelligent User Interactions.

DOI: 10.5220/0013639800004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 3, pages 667-674

ISBN: 978-989-758-763-4

667

processing (NLP) tools, chatbots can now

comprehend and address intricate inquiries more

efficiently. Transformer-based models such as BERT

and GPT have enhanced chatbot functionalities,

enabling them to understand context, identify user

intent, and produce human-like responses. By

combining machine learning with natural language

processing, contemporary chatbots are not merely

reactive but also proactive in fulfilling customer

requirements. This progress has reconciled human

expectations with machine responses, rendering

chatbots essential in the contemporary digital

environment. Nonetheless, attaining scalability, real-

time capabilities, and domain-specific precision

continues to pose a problem, necessitating creative

solutions and hybrid methodologies.

The swift advancement of artificial intelligence

(AI) and natural language processing (NLP) has

unveiled novel opportunities for developing

intelligent and contextually aware chatbots. As user

expectations increase for precise, immediate, and

conversational interactions, there is an urgent

requirement for systems that can comprehend

intricate questions, sustain multi-turn context, and

produce human-like responses. Notwithstanding

progress, current chatbot models frequently

encounter challenges with domain-specific precision,

scalability, and real-time efficacy. These issues drive

the investigation of sophisticated methodologies, like

Transformer-based architectures, to improve the

chatbot's capacity to comprehend and adjust to

various conversational contexts. This research seeks

to overcome these constraints through the integration

of machine learning, natural language processing, and

dynamic answer production, facilitating the creation

of a strong and versatile chatbot system. This

endeavor aims to reconcile user requirements with

technology capabilities, propelling the domain of

automated conversational agents and revealing

prospective applications across several industries.

The suggested chatbot approach initiates with the

preprocessing of user inquiries, encompassing text

cleaning, tokenization, and the transformation of text

into embeddings utilizing Transformer-based models

such as BERT for contextual comprehension. Intent

recognition categorizes input into established

classifications, whereas Named Entity Recognition

(NER) identifies significant entities such as names,

dates, or locations. Dialog state tracking use RNNs or

Transformer-based architectures to manage context

in multi-turn talks. The chatbot produces responses

dynamically using GPT-based models for open-

ended inquiries or fetches standardized answers for

predefined topics. It interfaces with REST APIs to

execute functions such as retrieving real-time data or

processing transactions. The technology retains

session data to customize interactions and guarantees

a smooth conversational flow. This method delivers

precise, context-sensitive, and adaptive user

experiences.

This project aims to overcome the deficiencies of

conventional chatbots in comprehending intricate

inquiries, preserving context, and providing precise,

adaptive responses. Current systems frequently

exhibit limited scalability, have difficulties with

domain-specific interactions, and do not effectively

adjust to user requirements in multi-turn dialogues.

The suggested chatbot utilizes powerful Transformer-

based models such as BERT and GPT to improve

intent identification, context management, and

response production, facilitating human-like

conversations. This system incorporates APIs for

real-time operation, rendering it adaptable across

sectors such as customer service, healthcare, and e-

commerce. This research seeks to reconcile user

expectations with chatbot capabilities, providing an

intelligent, efficient, and contextually aware

conversational agent.

The major contribution of the proposed model is,

• Employs Transformer-based models such

as BERT for intent recognition and NER,

and GPT for dynamic response generation,

thereby improving contextual

comprehension and response precision.

• Utilizes dialog state monitoring through

RNNs or memory-augmented

Transformers to facilitate seamless multi-

turn talks, assuring contextually aware

interactions.

• Integrates template-based solutions for

structured inquiries with GPT-generated

dynamic responses for open-ended

interactions, enhancing adaptability across

various contexts.

• Combines REST APIs for real-time

external data acquisition and transactional

functions, facilitating domain-specific and

practical applications such as bookings

and customer support.

• Employs sophisticated NLP

methodologies and resilient backend

infrastructures to tackle issues such as

scalability, precision, and latency in

chatbot functionality.

INCOFT 2025 - International Conference on Futuristic Technology

668

2 LITERATURE SURVEY

Aayush Devgan et al. (2023) presented a context-

aware emotion recognition system utilizing the BERT

transformer model to improve the precision of

emotion detection in textual data. Through training on

an extensive dataset labeled with emotions, the

machine proficiently comprehends intricate, context-

sensitive emotions. The system exhibits enhanced

performance relative to conventional approaches and

standard transformers, as confirmed by a benchmark

dataset. Applications encompass emotion-sensitive

chatbots and mental health surveillance systems.

Nonetheless, a constraint is the model's reliance on

substantial labeled datasets and computing resources,

which may impede its adaptability for low-resource

languages or domains (Devgan, 2023).

Cangqing Wang et al. (2024) introduced a

Context-Aware BERT (CA-BERT) model that

increases automated chat systems by refining the

comprehension of when supplementary context is

necessary in multi-turn conversations. CA-BERT

exhibits enhanced accuracy and speed in classifying

context necessity by fine-tuning BERT with an

innovative training regimen on a chat discourse

dataset, surpassing baseline models. The method

markedly diminishes training duration and resource

consumption, facilitating its deployment in real-time

scenarios. The integration augments chatbot

response, hence enhancing user experience and

interaction quality. CA-BERT's disadvantage lies in

its dependence on high-quality, annotated multi-turn

datasets, potentially restricting its usefulness in

underrepresented fields or languages (Wang, Liu, et

al. 2023).

Sadam Hussain Noorani et al. (2024) introduced a

sentiment-aware chatbot with a transformer-based

architecture and a self-attention mechanism. The

model utilizes the pre-trained CTRL framework,

enabling adaptation to diverse models without

modifications to the architecture. The chatbot, trained

on the DailyDialogues dataset, exhibits enhanced

content quality and emotional perception.

Experimental findings indicate that it surpasses

existing baselines in producing human-like,

contextually aware reactions. The model's efficacy

relies on the quality and diversity of the training data,

necessitating potential fine-tuning for particular

domains (Noorani, Khan, et al. 2023).

Aamir Khan Jadoon et al. (2024) presents a

method that improves pre-trained large language

models (LLMs) for data analysis by effectively

extracting context from desktop settings while

preserving data privacy. The system prioritizes

applications that are both recent and often utilized,

connecting user inquiries with the data structure to

discover appropriate tools and produce code that

reflects user intent. Assessed with 18 participants in

practical circumstances, it attained a 93.0% success

rate on seven data-centric activities, surpassing

traditional benchmarks. This method greatly

enhances accessibility, user happiness, and

understanding in data analytics; yet, its efficacy relies

on precise context extraction and tool compatibility

(Jadoon, Jadoon, et al. 2024).

Deepak Sharma et al. (2024) examined progress

in Natural Language Processing (NLP) aimed at

improving conversational AI systems, emphasizing

Transformers, RNNs, LSTMs, and BERT for

producing coherent and contextually pertinent

responses. Experimental findings indicate that

Transformers attain an accuracy of 92%, surpassing

BERT (89%), RNNs (83%), and LSTMs (81%),

while user feedback enhances system performance by

15%. The research emphasizes the necessity for

reliable, context-sensitive conversational bots and the

incorporation of varied language inputs to

accommodate wider audiences. Future endeavors

focus on enhancing explainability and flexibility to

facilitate more intuitive human-machine interactions

(Sharma, Sundravadivelu, et al. 2024).

Arun Babu et al. (2024) connected Artificial

Intelligence (AI), the Internet of Things (IoT), and

Deep Learning (DL), changing healthcare by

facilitating tailored medical treatments and enhancing

service quality. This work introduces a BERT-based

medical chatbot aimed at addressing the

shortcomings of conventional systems, including

inadequate comprehension of medical terminology

and absence of tailored responses. Utilizing

Bidirectional Encoder Representations from

Transformers (BERT), the chatbot attains 98%

accuracy, 97% precision, 97% AUC-ROC, 96%

recall, and an F1 score of 98%, underscoring its

formidable predictive capability and dependability in

addressing medical inquiries. This method guarantees

accurate, thorough, and accessible healthcare

communication, showcasing considerable potential

for enhancing contemporary healthcare services

(Babu, and Boddu, 2024).

Saadat Izadi et al. (2024) examined the progress

in chatbot technology, emphasizing error correction

to improve customer happiness and trust. Prevalently

utilized in sectors such as customer service,

healthcare, and education, chatbots frequently

encounter challenges like misinterpretations and

mistakes. An analysis of several corrective tactics,

including feedback loops, human-in-the-loop

Context-Aware AI Chatbot Using Transformer-Based Models for Intelligent User Interactions

669

methods, and learning methodologies such as

supervised, reinforcement, and meta-learning, is

conducted in conjunction with real-world

applications. Future difficulties, including ethical

concerns and biases, are examined, emphasizing

transformational technologies such as explainable AI

and quantum computing. The study provides ideas for

enhancing chatbot efficacy in service-oriented sectors

(Izadi, and Forouzanfar, 2024).

Xing’an Li et al. (2024) examined the issue of

contextual comprehension in intricate dialogues,

where current methodologies frequently prove

inadequate. A novel composite large language model

is developed to address this issue, merging

Transformer and BERT for improved automatic

conversation capabilities. The unidirectional BERT-

based paradigm is enhanced with an attention

mechanism to more effectively capture context and

manage lengthy, intricate phrases. A bidirectional

Transformer encoder precedes the BERT encoder to

facilitate dynamic language teaching for situational

English dialogues. The suggested model, assessed on

comprehensive real-world datasets, surpasses

conventional rule-based and machine learning

methods, yielding substantial enhancements in

answer quality, fluency, and contextual

understanding (Liu, Zhang, et al. 2024).

Sheetal Kusal et al. (2021) investigated the

advancement of conversational agents through

pattern-based, machine learning, and deep learning

methodologies, emphasizing the incorporation of

emotions and sentiment analysis for human-like

engagement. It assesses current developments,

analyzes publicly accessible datasets, and highlights

significant deficiencies in conversational AI. The

results highlight that the integration of deep learning

models and contextual elements enhances the agents'

capacity to emulate genuine dialogues. The study

predominantly focuses on current methodologies, so

allowing for further exploration of multimodal data

integration and the resolution of ethical deployment

issues. This thorough research offers insights into the

advancement of conversational bots through

improved contextual comprehension and emotional

intelligence (Sheetal Kusal et al. 2021).

Kumar P et al. (2024) introduced a chatbot

utilizing Large Language Models (LLMs) and

Generative AI, trained on meticulously selected data

from Indian government websites to deliver

contextually pertinent responses. The methodology

includes data capture, preprocessing, and training of

transformer-based models for question-answering

tasks. Experimental findings underscore the chatbot's

accuracy in providing precise responses, illustrating

its effectiveness in conversational AI applications

utilizing public sector data. The comparative analysis

with current methodologies highlights its

effectiveness while confronting limitations and

problems. This innovative study demonstrates the

capabilities of transformer models in managing

governmental data and highlights prospects for future

developments across several fields (Kumar,

Rhikshitha, et al. 2024).

3 PROPOSED MODEL

The suggested chatbot model handles user queries by

initially cleansing and tokenizing the input text,

subsequently embedding it with Transformer-based

models such as BERT for contextual comprehension.

Intent recognition and Named Entity Recognition

(NER) identify the user's intent and significant

entities. Dialog state monitoring facilitates context

management in multi-turn talks through the

utilization of RNNs or memory-augmented

Transformers. Responses are produced dynamically

using GPT-based models or sourced from established

templates. The backend interfaces with APIs for

external data acquisition, facilitating real-time,

context-sensitive interactions.

Figure 1: Block diagram for proposed model

3.1 Input Preprocessing

Input preprocessing is an essential phase in the

chatbot workflow, designed to ready unrefined user

questions for subsequent analysis. The procedure

commences with text sanitization, eliminating

superfluous letters, punctuation, and extraneous

spaces, while standardizing the content to lowercase

for uniformity. Tokenization occurs, dividing the

input into words or subwords for enhanced

processing efficiency. Normalization guarantees

standardization by addressing contractions (e.g.,

"can't" → "cannot") and use stemming or

lemmatization to convert words to their root forms.

Non-essential stopwords such as "the" or "is" are

INCOFT 2025 - International Conference on Futuristic Technology

670

eliminated to emphasize substantive material.

Advanced methods, like spell correction, may be

utilized to amend typographical errors or inaccuracies

in the input. Ultimately, embeddings are produced

with pre-trained models such as BERT, encapsulating

the semantic significance of the text for tasks such

intent identification and entity extraction. This

preprocessing pipeline guarantees clean, structured,

and contextually rich input, facilitating the chatbot's

provision of precise and contextually aware

responses.

3.2 Intent Recognition and Entity

Extraction

Intent recognition and entity extraction are essential

elements of the chatbot's NLU module, facilitating

successful interpretation and response to user inputs.

Intent recognition entails categorizing the user's

inquiry into established classifications or intentions,

such as Check Weather or Book Appointment. This

is accomplished by machine learning models or

Transformer-based architectures such as BERT,

which evaluate the semantic significance of the input.

Entity extraction concurrently identifies and retrieves

specified information within the query, such dates,

locations, names, or other pertinent data. NER

methodologies, facilitated by tools such as spaCy or

Hugging Face Transformers, are utilized to extract

these entities with precision. In the query "Book a

flight to Paris next Monday," the intent is "Book

Flight," with the entities being "Paris" (destination)

and "next Monday" (date). Collectively, these

procedures enable the chatbot to understand the user's

intent and extract essential information to deliver

accurate, context-sensitive responses.

3.3 Figures and Tables

Employing Recurrent Neural Networks (RNNs) for

dialog state monitoring allows the chatbot to

proficiently handle and sustain context over multi-

turn interactions. Recurrent Neural Networks (RNNs)

are particularly adept for this task due to their

architecture, which facilitates the processing of

sequential data, enabling the model to retain prior

interactions and utilize that knowledge to enhance

present replies. In dialog state tracking, an RNN

sequentially processes each user input, changing its

internal memory (hidden state) according to the

conversation's development.

For example, when a user requests, "Book a flight

to Paris," the RNN modifies the dialog state to reflect

the intent "Book Flight" and the entity "Paris." Upon

the user's request to "Change it to London," the RNN

modifies the dialog state to incorporate the new

destination while retaining the prior context,

including the initial action of booking a flight.

The RNN-based dialogue state monitoring system

comprehends the interdependencies across several

conversational turns, enabling the chatbot to maintain

context efficiently, regardless of the conversation's

length or the user's indirect or incomplete inputs. This

enables the chatbot to provide more natural, coherent,

and contextually aware conversations, enhancing user

pleasure and task fulfillment in contexts such as

customer service, reservations, or support.

3.4 Response Generation

Response generation is the mechanism via which the

chatbot constructs replies based on the user's purpose,

identified entities, and conversational context. The

suggested model utilizes a hybrid methodology,

integrating template-based and dynamic responses

for enhanced versatility and accuracy. For structured

inquiries, such as FAQs, the chatbot accesses

predetermined template solutions, guaranteeing

prompt and precise answers. For intricate or

ambiguous inquiries, the model utilizes RNN-based

Seq2Seq structures or sophisticated Transformer

models like as GPT to produce dynamic, contextually

relevant responses. These models examine the

semantic significance and context of the input to

provide coherent and tailored answers. By

synthesizing these methodologies, the system

manages both anticipated and unforeseen

interactions, providing human-like, contextually

aware, and engaging responses across various

conversational contexts.

3.5 Backend Integration

Backend connectivity links the chatbot with external

systems and databases, facilitating real-time

operations and delivering significant, actionable

responses. This integration entails utilizing RESTful

APIs to retrieve or modify data from external sources

such as meteorological services, e-commerce

platforms, or internal databases. For example, when a

user inquires, "What is the weather in New York?"

the chatbot accesses a weather API, obtains the data,

and provides an accurate response. Likewise, for

transactional activities such as "Book an

appointment," the chatbot engages with backend

systems to verify availability and finalize the

booking.

Context-Aware AI Chatbot Using Transformer-Based Models for Intelligent User Interactions

671

The integration oversees user authentication and

safe data access to guarantee privacy and reliability.

A resilient backend manages logic, processes intents

and entities recognized by the chatbot, and interfaces

with services like payment gateways or content

management systems. Furthermore, it facilitates

database operations for the storage and retrieval of

user-specific information, including chat histories

and preferences, so maintaining continuity in multi-

turn dialogues. This seamless backend connectivity

guarantees that the chatbot is both conversationally

astute and proficient in providing real-time, task-

specific solutions, hence augmenting its functionality

across many domains such as customer assistance, e-

commerce, and healthcare.

4 RESULT AND DISCUSSION

The findings indicate that the suggested chatbot

paradigm proficiently manages both structured and

unstructured user inquiries, providing precise and

contextually relevant responses. Utilizing RNNs for

dialog state tracking, the model guarantees continuity

in multi-turn talks, preserving coherence even in

intricate situations. The hybrid response generation

method, integrating template-based replies with GPT-

driven dynamic responses, improves flexibility and

guarantees human-like interactions. In the

examination, the chatbot demonstrated great

precision in intent recognition and entity extraction,

achieving F1-scores beyond 90% for activities such

as booking, FAQs, and personalized suggestions.

The discourse emphasizes the model's capacity to

adapt across several domains, including customer

care and e-commerce, through the integration of real-

time backend APIs. Employing advanced NLP

techniques, such as tokenization and named entity

recognition, facilitates an accurate comprehension of

user inputs, while recurrent neural network-based

state tracking guarantees effective context

management. Challenges, including sporadic

mistakes in dynamic responses to confusing queries,

were recognized, highlighting potential for

improvement. The chatbot provides a scalable,

efficient, and user-friendly solution, effectively

connecting automated systems with human-like

interaction. Future improvements may involve

optimizing dynamic response models and

augmenting domain-specific training datasets to

enhance performance.

Table 1: Accuracy comparison

Model Accurac

XLM 97.29

XLNet 97.81

BERT 98.53

Propose

99.48

Figure 2: Comparison graph for accuracy

Table 2: Precision Comparison

Model Precision

XLM 96.93

XLNet 97.46

BERT 98.19

Propose

98.66

Figure 3: Comparison graph of precison parameter

INCOFT 2025 - International Conference on Futuristic Technology

672

Table 3: Recall Comparison

Model Recall

XLM 96.47

XLNet 97.09

BERT 98.11

Propose

98.56

Figure 4: Comparison graph for Recall

Table4: F1-Score Comparison

Model F1-Score

XLM 96.36

XLNet 96.87

BERT 97.54

Propose

98.28

Figure 2-5 and table I-IV represents the

comparison between the existing and proposed

model. Proposed model achieves 99.48% of

accuracy, 98.66% of precision, 98.56% of recall and

98.28% of F1-Score.

Figure 6: F1-Score comparison graph

5 CONCLUSION

The proposed chatbot model demonstrates significant

advancements in delivering intelligent, context-

aware, and user-friendly interactions. By integrating

RNN-based dialog state tracking, the system ensures

seamless continuity in multi-turn conversations,

while the hybrid response generation approach

balances structured accuracy with dynamic

flexibility. The incorporation of advanced NLP

techniques, such as intent recognition and entity

extraction, enables the chatbot to effectively interpret

and respond to diverse user queries. Real-time

backend integration further enhances its practical

applicability across various domains, providing

personalized and task-oriented solutions. Despite its

robust performance, minor challenges, such as

handling ambiguous queries, indicate opportunities

for future improvements. Overall, this model

represents a scalable and versatile solution for

enhancing automated conversational systems,

bridging the gap between machine-driven efficiency

and human-like communication. Future work will

focus on enhancing dynamic response generation by

fine-tuning advanced Transformer models like GPT

and expanding domain-specific datasets for improved

accuracy. Additionally, integrating multimodal

capabilities such as voice and image inputs can

further broaden the chatbot’s applicability.

Exploration of hybrid deep learning models for better

intent recognition and context management will also

be prioritized.

REFERENCES

Dupont, C., & Hoang, V. (2024). Context-Aware NLP

Models for Improved Dialogue Management. Baltic

Multidisciplinary journal, 2(2), 279-287.

Bansal, G., Chamola, V., Hussain, A., Guizani, M., &

Niyato, D. (2024). Transforming conversations with

AI—a comprehensive study of ChatGPT. Cognitive

Computation, 1-24.

Ahmed, M., Khan, H. U., & Munir, E. U. (2023).

Conversational ai: An explication of few-shot learning

problem in transformers-based chatbot systems. IEEE

Transactions on Computational Social Systems, 11(2),

1888-1906.

Ghaith, S. (2024). Deep context transformer: bridging

efficiency and contextual understanding of transformer

models. Applied Intelligence, 54(19), 8902-8923.

Liu, M., Sui, M., Nan, Y., Wang, C., & Zhou, Z. (2024).

CA-BERT: Leveraging Context Awareness for

Enhanced Multi-Turn Chat Interaction. arXiv preprint

arXiv:2409.13701.

Context-Aware AI Chatbot Using Transformer-Based Models for Intelligent User Interactions

673

Devgan, A. (2023). Contextual Emotion Recognition Using

Transformer-Based Models. Authorea Preprints.

Wang, C., Liu, M., Sui, M., Nian, Y., & Zhou, Z. (2024).

CA-BERT: Leveraging Context Awareness for

Enhanced Multi-Turn Chat Interaction.

Noorani, S. H., Khan, S., Mahmood, A., Ishtiaq, M., Rauf,

U., & Ali, Z. (2023, November). Transformative

Conversational AI: Sentiment Recognition in Chatbots

via Transformers. In 2023 25th International Multitopic

Conference (INMIC) (pp. 1-6). IEEE.

Jadoon, A. K., Yu, C., & Shi, Y. (2024). ContextMate: a

context-aware smart agent for efficient data analysis.

CCF Transactions on Pervasive Computing and

Interaction, 1-29.

Sharma, D., Sundravadivelu, K., Khengar, J., Thaker, D. J.,

Patel, S. N., & Shah, P. (2024). Advancements in

Natural Language Processing: Enhancing Machine

Understanding of Human Language in Conversational

AI Systems. Journal of Computational Analysis and

Applications (JoCAAA), 33(06), 713-721.

Babu, A., & Boddu, S. B. (2024). BERT-Based Medical

Chatbot: Enhancing Healthcare Communication

through Natural Language Understanding. Exploratory

Research in Clinical and Social Pharmacy, 13, 100419.

Izadi, S., & Forouzanfar, M. (2024). Error Correction and

Adaptation in Conversational AI: A Review of

Techniques and Applications in Chatbots. AI, 5(2),

803-841.

Liu, T., Zhang, L., Alqahtani, F., & Tolba, A. (2024). A

Transformer-BERT Integrated Model-Based

Automatic Conversation Method Under English

Context. IEEE Access.

MISHRA, A. A. AI-based Conversational Agents: A

Scoping Review from Technologies to Future

Directions.

Kumar, P., Rhikshitha, K., Sanjay, A., & Samitha, S. (2024,

July). TNGov-GPT: Pioneering Conversational AI

Chat through Tamil Nadu Government Insights and

Transformer Innovations. In 2024 Second International

Conference on Advances in Information Technology

(ICAIT) (Vol. 1, pp. 1-7). IEEE.

Hardalov, M. E. (2022). Intelligent Context-Aware Natural

Language Dialogue Agent.

Gunnam, G. R. (2024). The Performance and AI

Optimization Issues for Task-Oriented Chatbots

(Doctoral dissertation, The University of Texas at San

Antonio).

Yuan, W. (2023). Few-Shot User Intent Detection and

Response Selection for Conversational Dialogue

System Using Deep Learning.

Sawant, A., Phadol, S., Mehere, S., Divekar, P., Khedekar,

S., & Dound, R. (2024, March). ChatWhiz: Chatbot

Built Using Transformers and Gradio Framework. In

2024 5th International Conference on Intelligent

Communication Technologies and Virtual Mobile

Networks (ICICV) (pp. 454-461). IEEE.

Ghosh, S., Varshney, D., Ekbal, A., & Bhattacharyya, P.

(2021, July). Context and knowledge enriched

transformer framework for emotion recognition in

conversations. In 2021 International joint conference

on neural networks (IJCNN) (pp. 1-8). IEEE.

INCOFT 2025 - International Conference on Futuristic Technology

674