Unifying Large Language Models and Knowledge Graphs: A Roadmap

IEEE Transactions on Knowledge and Data Engineering, Год журнала: 2024, Номер 36(7), С. 3580 - 3599

Опубликована: Янв. 10, 2024

Large language models (LLMs), such as ChatGPT and GPT4, are making new waves in the field of natural processing artificial intelligence, due to their emergent ability generalizability. However, LLMs black-box models, which often fall short capturing accessing factual knowledge. In contrast, Knowledge Graphs (KGs), Wikipedia Huapu for example, structured knowledge that explicitly store rich KGs can enhance by providing external inference interpretability. Meanwhile, difficult construct evolve nature, challenges existing methods generate facts represent unseen Therefore, it is complementary unify together simultaneously leverage advantages. this article, we present a forward-looking roadmap unification KGs. Our consists three general frameworks, namely, 1) KG-enhanced LLMs, incorporate during pre-training phases LLMs, or purpose enhancing understanding learned LLMs; xmlns:xlink="http://www.w3.org/1999/xlink">2) LLM-augmented KGs, different KG tasks embedding, completion, construction, graph-to-text generation, question answering; xmlns:xlink="http://www.w3.org/1999/xlink">3) Synergized + KGs , play equal roles work mutually beneficial way both bidirectional reasoning driven data We review summarize efforts within these frameworks our pinpoint future research directions.

Язык: Английский

---

When large language models meet personalization: perspectives of challenges and opportunities

World Wide Web, Год журнала: 2024, Номер 27(4)

Опубликована: Июнь 28, 2024

Abstract The advent of large language models marks a revolutionary breakthrough in artificial intelligence. With the unprecedented scale training and model parameters, capability has been dramatically improved, leading to human-like performances understanding, synthesizing, common-sense reasoning, etc. Such major leap forward general AI capacity will fundamentally change pattern how personalization is conducted. For one thing, it reform way interaction between humans systems. Instead being passive medium information filtering, like conventional recommender systems search engines, present foundation for active user engagement. On top such new foundation, users’ requests can be proactively explored, required delivered natural, interactable, explainable way. another also considerably expand scope personalization, making grow from sole function collecting personalized compound providing services. By leveraging as general-purpose interface, may compile user’s into plans, calls functions external tools (e.g., calculators, service APIs, etc.) execute integrate tools’ outputs complete end-to-end tasks. Today, are still rapidly developed, whereas application largely unexplored. Therefore, we consider right time review challenges opportunities address them with models. In particular, dedicate this perspective paper discussion following aspects: development existing system, newly emerged capabilities models, potential ways use personalization.

Язык: Английский

---

Knowledge graph-based manufacturing process planning: A state-of-the-art review

Journal of Manufacturing Systems, Год журнала: 2023, Номер 70, С. 417 - 435

Опубликована: Авг. 24, 2023

Язык: Английский

---

Joint Knowledge Graph and Large Language Model for Fault Diagnosis and Its Application in Aviation Assembly

IEEE Transactions on Industrial Informatics, Год журнала: 2024, Номер 20(6), С. 8160 - 8169

Опубликована: Март 8, 2024

In complex assembly industry settings, fault localization involves rapidly and accurately identifying the source of a obtaining troubleshooting solution based on symptoms. This study proposes knowledge-enhanced joint model that incorporates aviation knowledge graph (KG) embedding into large language models (LLMs). utilizes graph-structured Big Data within KGs to conduct prefix-tuning LLMs. The for enable an online reconfiguration LLMs, which avoids massive computational load. Through subgraph learning process, specialized domain, especially in localization, is strengthened. context functional testing, can generate subgraphs, fuse through retrieval augmentation, ultimately provide knowledge-based reasoning responses. practical industrial scenario experiments, enhancement demonstrates accuracy 98.5% diagnosis schemes.

Язык: Английский

---

A systematic literature review of knowledge graph construction and application in education

Heliyon, Год журнала: 2024, Номер 10(3), С. e25383 - e25383

Опубликована: Фев. 1, 2024

In the dynamic landscape of modern education, search for improved pedagogical methods, enriched learning experiences, and empowered educators remains a perpetual pursuit. recent years, remarkable technological innovation has asserted its dominance in education: Knowledge Graphs (KGs). These structured representations knowledge are increasingly proving to be indispensable tools, fostering advancements driven by growing recognition their essential role enriching personalised learning, curriculum design, concept mapping, educational content recommendation systems. this paper, systematic literature review (SLR) been conducted comprehensively examine KG construction methodologies applications across five key domains education. each examined study, we highlight specific functionalities, extraction techniques, base characteristics, resource requirements, evaluation criteria, limitations. This paper distinguishes itself offering broad overview KGs analyzing state-of-the-art methodologies, identifying research gaps limitations, paving way future advancements.

Язык: Английский

---

Large language models as oracles for instantiating ontologies with domain-specific knowledge

Knowledge-Based Systems, Год журнала: 2025, Номер unknown, С. 112940 - 112940

Опубликована: Янв. 1, 2025

Язык: Английский

---

Augmenting general-purpose large-language models with domain-specific multimodal knowledge graph for question-answering in construction project management

Advanced Engineering Informatics, Год журнала: 2025, Номер 65, С. 103142 - 103142

Опубликована: Фев. 15, 2025

Язык: Английский

---

Prompt engineering of GPT-4 for chemical research: what can/cannot be done?

Science and Technology of Advanced Materials Methods, Год журнала: 2023, Номер 3(1)

Опубликована: Сен. 20, 2023

This paper evaluates the capabilities and limitations of Generative Pre-trained Transformer 4 (GPT-4) in chemical research. Although GPT-4 exhibits remarkable proficiencies, it is evident that quality input data significantly affects its performance. We explore GPT-4's potential tasks, such as foundational chemistry knowledge, cheminformatics, analysis, problem prediction, proposal abilities. While language model partially outperformed traditional methods, black-box optimization, fell short against specialized algorithms, highlighting need for their combined use. The shares prompts given to responses, providing a resource prompt engineering within community, concludes with discussion on future research using large models.

Язык: Английский

---

A large language model-enabled machining process knowledge graph construction method for intelligent process planning

Advanced Engineering Informatics, Год журнала: 2025, Номер 65, С. 103244 - 103244

Опубликована: Март 8, 2025

Язык: Английский

---

Actionable Cyber Threat Intelligence Using Knowledge Graphs and Large Language Models

Опубликована: Июль 8, 2024

Язык: Английский

---