Leveraging Knowledge Graphs and LLMs for Context-Aware Messaging PubID: pubid: 979-8-3315-2131-8/25/$31.00 © 2025 IEEE
1 st Rajeev Kumar, 2 nd Harishankar Kumar, 3 rd Kumari Shalini · 2025
Read Leveraging Knowledge Graphs and LLMs for Context-Aware Messaging PubID: pubid: 979-8-3315-2131-8/25/$31.00 © 2025 IEEE by 1 st Rajeev Kumar, 2 nd Harishankar Kumar, 3 rd Kumari Shalini (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-13499.txt.
Introduction
Rajeev Kumar, Harishankar Kumar, and Kumari Shalini proposed a messaging framework in twenty twenty-five that rewrites communications using knowledge graphs and large language models. Personalized messages land better in health, education, and hiring, but hand-tailoring does not scale. Their system stores people, places, and events as graph nodes and links a draft's mentions to those nodes.
Relevant facts — preferences, roles, cultural norms — are pulled from the graph and combined with the draft, and a language model generates the personalized version. Reported acceptance rates reached forty-two percent in healthcare, fifty-three percent in education, and seventy-eight percent in professional recruitment. The authors present entity linking plus language modeling as a structured, scalable route to audience-aware communication.
Catalog record
| CREATOR | 1 st Rajeev Kumar, 2 nd Harishankar Kumar, 3 rd Kumari Shalini |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-13499
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Leveraging Knowledge Graphs and LLMs for Context-Aware Messaging PubID: pubid: 979-8-3315-2131-8/25/$31.00 © 2025 IEEE 1 st Rajeev Kumar, 2 nd Harishankar Kumar, 3 rd Kumari Shalini Abstract Personalized messaging plays an essential role in improving communication in areas such as healthcare, education, and professional engagement. This paper introduces a framework that uses the Knowledge Graph (KG) to dynamically rephrase written communications by integrating individual and context-specific data. The knowledge graph represents individuals, locations, and events as critical nodes, linking entities mentioned in messages to their corresponding graph nodes. The extraction of relevant information, such as preferences, professional roles, and cultural norms, is then combined with the original message and processed through a large language model (LLM) to generate personalized responses. The framework demonstrates notable message acceptance rates in various domains: 42% in healthcare, 53% in education, and 78% in professional recruitment. By integrating entity linking, event detection, and language modeling, this approach offers a structured and scalable solution for context-aware, audience-specific communication, facilitating advanced applications in diverse fields. I Introduction Effective communication in domains such as healthcare, education, and professional recruitment is challenging due to the need for personalized and context-aware messaging. Generic messaging systems often fail to address individual preferences or adapt to real-time contexts, leading to reduced effectiveness. M.A. Meier [1] emphasizes the importance of domain-specific knowledge in crafting relevant messages and highlights the limitations of one-size-fits-all approaches across various fields. They note that understanding the context of the recipient and minimizing personal biases in message generation requires skilled training and deep domain knowledge. Johnson et al. (2020) [2] explored the impact of personalized messages in education, showing that aligning messages with students’ interests and learning styles improved engagement significantly. They demonstrated that interest-based communication increased student retention rates. Similarly, healthcare reminders that overlook real-time factors, such as adverse weather conditions, often result in missed appointments, underlining the need for adaptable and context-sensitive systems. This work introduces a Context-Aware Messaging System that bridges these gaps by integrating three key components: knowledge graphs, real-time event data, and large language models. Wang et al. (2022) [3] explored the potential of combining structured knowledge representation with conversational AI, showing its impact on the generation of contextually accurate responses. Building on this, the proposed system leverages knowledge graphs to represent entities such as individuals, locations, and events, dynamically linking contextual data to personalize messages. Real-time factors, like weather updates or local events, are incorporated to enhance the relevance of generated messages. By aligning communication with individual preferences and situational contexts, this system addresses the shortcomings of traditional approaches, offering a scalable and intelligent solution across critical domains. II Literature Review The field of personalized messaging has gained significant attention due to its potential to enhance engagement across diverse domains. Existing research highlights various approaches to improve user interest relevance, including the use of knowledge graphs (KG), real-time data integration, and advances in natural language processing (NLP) with LLM. This section reviews related work and outlines the gaps addressed by the proposed system. II- 1 Knowledge Graphs for Personalization Knowledge graphs (KGs) are a powerful way to personalize user experiences. By organizing information into connected networks of entities and their relationships, they help systems understand and respond better to individual needs and preferences. When combined with large language models (LLMs), KGs can make personalization even more detailed and relevant to the context. For example, KGs have been used to improve recommendations by recording user preferences and product details (Wang et al., 2019) [4]. In healthcare, they help create personalized treatment plans by combining patient data, medical knowledge, and clinical guidelines (Shi et al., 2021) [5]. In education, KGs help customize learning materials for individual students (Yu et al., 2022) [6]. Recent progress has focused on integrating real-time event data with KGs and LLMs to build adaptive systems. For example, Shen et al. (2023) [7] showed how adding time-related information to KGs can improve recommendations, and Wang et al. (2022) [8] demonstrated how KGs can improve conversational AI by providing better context. However, most of these efforts have been limited to search, recommendations, or data analysis. There is still a noticeable gap in using KGs for real-time personalized message writing. Filling this gap could lead to systems that deliver highly customized and meaningful messages, transforming how users engage and interact. II- 2 Real-Time Context in Messaging: State-of-the-Art Approaches Real-time context in messaging systems has shown notable promise in fields such as healthcare and education. For example, Lee et al. (2021) [9] reported a 10% increase in patient attendance by using weather-based appointment reminders, while Johnson et al. (2020) [10] boosted assignment completion rates by 15% through context-aware notifications for students. However, many such methods do not fully integrate live events with user profiles for a deeper personalization. Recent work combining knowledge graphs (KGs), large language models (LLMs), and real-time data, for example, Smith et al. [23] with an acceptance rate of 40% in healthcare reminders and Zhang et al. [24] reaching 50% engagement in educational messaging, still relies heavily on static user information or focuses narrowly on specific domains. To address these shortcomings, our proposed framework continuously updates KGs with fresh event data (e.g., local disruptions, schedule changes) and employs a fine-tuned LLM to dynamically tailor messages. By unifying real-time event extraction, KG-driven contextual data, and advanced language modeling, we aim to surpass existing approaches in delivering adaptive, context-rich communication for healthcare, education, and recruitment. II- 3 Large Language Models in Context-Aware Messaging Large Language Models (LLMs) like GPT and BERT have transformed natural language processing by enabling advanced text generation and understanding. Brown et al. (2020) [11] showed how LLMs can produce coherent and context-aware text, which has been successfully used in areas such as customer support and content creation. More recently, Sun et al. (2023) [12] explored combining LLMs with structured data sources like knowledge graphs (KGs) to improve the accuracy of context-aware responses. However, these methods often treat KG and LLM as separate tools, without fully integrating them. The proposed system addresses this limitation by incorporating the context retrieved from KG directly into LLM prompts, ensuring that the generated messages are both highly relevant to the context and tailored to the audience. II- 4 Gaps and Opportunity in Existing System Despite significant progress in personalized communication, existing systems face several limitations that hinder their effectiveness. One major challenge is entity disambiguation, particularly when user profiles are ambiguous or incomplete. This often leads to inaccurate or irrelevant messages. Additionally, the use of real-time data is typically restricted to narrow applications, which limits the flexibility and broader applicability of these systems. Another gap is the lack of robust feedback integration mechanisms. Many current approaches do not incorporate user feedback into their workflows, which is crucial to refining and improving system performance over time. These limitations create an opportunity for innovation to build more adaptive and scalable solutions. The proposed Context-Aware Messaging System addresses some of these gaps by integrating knowledge graph-based personalization, real-time event data, and fine-tuned LLMs into a unified framework. This system will also include in future iteration mechanisms for collecting and leveraging user feedback through active learning, enabling continuous improvement and adaptability across diverse domains. This holistic approach positions the