Summary and Objectives

Summary

Significant progress has been made recently in Natural Language Processing (NLP) due to the development of Large Language Models (LLMs) and
their capacity to understand and produce text together with their capability to successfully address different applications. Despite the fact that the
number, size and capabilities of LLMs are continuously growing, they still present a number of limitations and challenges, not only related to their
computational cost and training data needs, but also to their performance in solving some linguistic phenomena and tasks that can highly affect the trust and reliability of AI systems. Due to these limitations and in parallel to this great LLM development, other emerging research areas are starting to gain traction. One of these emerging areas is related to applying quantum mechanics principles to the field of AI, and more specifically Quantum NLP (QNLP).
The motivation for merging quantum computing and NLP lies in quantum systems ability to handle complex, high-dimensional data more efficiently. On the one hand, quantum computers could provide exponential speedups in tasks like semantic analysis, summarization, text classification and word embeddings generation, which are computationally expensive on classical systems. On the other hand, quantum algorithms could facilitate more advanced language models and techniques to grasp subtle nuances in textual data, as well as enable more efficient processing of long texts, improving the ability to condense information quickly. However, this is still largely theoretical, still in its nascent stages, as it has yet to be better proven with practical implementations.
Our project, QUantum Mechanics for LAnguage Understanding anD gEneration (QUMLAUDE), seeks to investigate how quantum information
representation and algorithms can address complex NLP tasks such as text simplification and summarization. By leveraging quantum-inspired
algorithms, we aim to model nuanced linguistic phenomena with special attention to semantic representationsuch as contextual embeddings, semantic composition, and metaphorswhere quantum advancements promise to deliver more accurate and precise results compared to current methods.
To successfully meet the projects objective, the research work has been organized into six modules, with a duration of 3 years: Quantum representation of the text (M1); Quantum and Quantum-inspired Algorithms for NLP (M2); Integrating quantum theory in NLP tasks and applications (M3); Use Case: Retrieval Augmented Generation system for the innovation domain (M4); Dissemination (M5) and project management (M6).
QUMLADUE project brings together a multidisciplinary team of experts from linguistics, computational sciences, and experimental sciences, positioning the project at the forefront of innovation in quantum computing. Our research will not only deepen our understanding of quantum-enhanced NLP but also foster new standards in quantum computing applications.

Objectives

The primary objective of the project is to investigate how quantum information representation and quantum algorithms can contribute to complex NLP tasks related to language understanding and generation (e.g., text simplification and summarization). This includes exploring quantum-inspired algorithms capable of modeling linguistic phenomena, with a particular focus on semantic representation (e.g., contextual embeddings, semantic composition, metaphors) and other areas where quantum advancements could provide significant improvements. The aim is to achieve more precise and accurate results compared to current approaches, while making them more interpretable, transparent, and therefore, reliable.

To achieve this goal, the following specific objectives are defined:

  • OB1. To analyse and establish how to apply quantum theories to the representation of texts identifying current limitations and challenges.
  • OB2. To study the state of the art on available quantum algorithms and frameworks for NLP.
  • OB3. To explore different optimisation mechanisms that can be used to propose either pure quantum models or hybrid models combined with LLMs and compare them.
  • OB4. To research, propose and develop novel NLP approaches that integrate the principles of quantum information theory for language understanding and generation tasks (e.g., text simplification and summarisation) and the linguistic phenomena involved in them related to the
    semantic representation of text, such as contextual embeddings, knowledge graphs, word sense disambiguation, anaphora resolution, or metaphor interpretation.
  • OB5. To design and development of a specialised RAG (Retrieval-Augmented Generation) system for retrieving and generating information on innovation. This system will integrate advancements in QNLP tasks developed within the project, such as contextual embeddings, semantic
    composition, knowledge graphs, anaphora resolution, metaphors, and other areas where quantum advancements can provide significant improvements. This use case will serve as an experimental platform and testing ground to evaluate intrinsically and extrinsically the impact and effectiveness of the developments in NLP, ensuring a practical and results-oriented approach.
  • OB6. To promote and disseminate the research results obtained from the project through different national and international media—including well-indexed journals, conferences, seminars, etc.—as well as exploit the potential for transferring this technology to society.