June 26, 2024

Haltia.AI Unveils Ontology-Driven Symbolic Knowledge Capture Study

Code And Datasets Available On GitHub

Haltia.AI, a personal AI startup, has released an AI research paper titled ‘Prompt-Time Ontology-Driven Symbolic Knowledge Capture with Large Language Models.’

The study heralds a significant paradigm shift in AI personalisation, introducing a groundbreaking methodology that leverages ontology and knowledge-graph frameworks to transform user interactions with AI applications.

The cornerstone of Haltia.AI’s research lies in the fusion of ontology-driven symbolic knowledge capture with generative AI, bridging the gap between natural language processing and factual reasoning.

By integrating the symbolic power of knowledge graphs with the linguistic prowess of large language models (LLMs), the company aims to imbue AI systems with unprecedented intelligence and responsiveness, heralding a new era of personalised AI interactions.

Arto Bendiken, Chief Technology Officer of Haltia.AI, underlined the potential impact of this breakthrough: “Our integration of ontology-driven knowledge capture with LLMs represents a monumental leap in AI personalisation.

“By infusing AI systems with the ability to comprehend and utilise predefined ontologies, we empower them to deliver tailored, contextually relevant responses to user queries—a game-changer for the AI landscape.”

Predefined Ontologies

Central to the research is developing methodologies enabling LLMs to grasp and leverage predefined ontologies effectively. Utilising the KNOW ontology, designed to model personal information, facilitates the structured extraction of user data, enhancing accuracy and relevance.

Dr Tolga Çöplü is the lead author of ‘Prompt-Time Ontology-Driven Symbolic Knowledge Capture with Large Language Models.’ Credit: Haltia.AI

Unlike conventional in-context learning methods, Haltia.AI’s approach employs fine-tuning—a more practical and scalable technique—to enable the language model to internalise the ontology and streamline data processing, instilling confidence in the research’s practicality.

Dr Tolga Çöplü, the paper’s lead author, underscored the research’s real-world implications: “Our findings pave the way for AI systems capable of capturing and utilising personal knowledge with unparalleled precision.

“Through the synergistic fusion of neural networks and symbolic AI via fine-tuning, we offer a scalable and efficient solution poised to revolutionise diverse AI applications.”

The study showcases promising results, demonstrating high success rates with a modest number of training samples.

Performance evaluations underscore the efficacy of fine-tuning with a diverse training dataset in augmenting the model’s capacity for ontology-based knowledge capture.

Furthermore, in a gesture of commitment to transparency and collaboration, Haltia.AI has made the code and datasets in the study publicly accessible on GitHub, fostering an open-source ethos within the AI community.

Crete Presentation

Haltia.AI unveiled its paper at the Extended Semantic Web Conference 2024 (May 26-30) in Crete, Greece.

This upcoming study will delve deeper into the KNOW ontology, showcasing the company’s prowess in data modelling and knowledge engineering—an unprecedented feat in the industry.

Haltia.AI continues solidifying its leadership position in shaping the future of practical AI solutions by championing the cutting-edge approach of neuro-symbolic synthesis.

Haltia.AI’s latest breakthrough represents a significant stride forward in personalised AI companions. By harnessing the power of ontology-driven symbolic knowledge capture, this study offers a holistic solution for enhancing AI personalisation capabilities, setting a new benchmark for future research endeavours.

Haltia.AI’s commitment to open research is commendable, particularly for a startup of its size and stage. Its dedication to fostering transparency and collaboration underscores a rare blend of innovation and altruism, positioning the company as a frontrunner in the global AI landscape.

For additional insights and access to the code and datasets, interested parties are encouraged to visit the Haltia.AI GitHub Repository.

Featured image: Haltia.AI’s latest breakthrough represents a significant stride forward in personalised AI companions. Credit: Growtika

    Middle East News 247

    Middle East News 247 delivers trending business and lifestyle news and essential infotainment for, and from the Middle East region, with key focus on the GCC nations: United Arab Emirates (UAE), Saudi Arabia, Qatar, Bahrain, Kuwait, and Oman.
    Follow Me:

    Related Posts

    Leave a Reply

    Your email address will not be published. Required fields are marked *