Abstract
dc:descriptionThis thesis investigates the application of a quantum natural language processing model to extractive text summarization. The aim is to evaluate the applicability of a QNLP model built with lambeq to a sentence classification task and to identify practical limitations of this approach. In this work, extractive summarization is formulated as a binary sentence classification problem, where each sentence is classified as either suitable or not suitable for inclusion in a summary. A sentence-level dataset was constructed using WikiHow texts and weakly supervised labels generated with a pre-trained extractive summarization model. The sentences were syntactically simplified, transformed into grammatical diagrams, and then converted into parameterized quantum circuits. The model was trained using QuantumTrainer and NumpyModel from lambeq library. An initial experiment with TketModel and AerSimulator was also performed, showing that the shot-based execution scenario was unstable for this task. The experiments evaluated several model configurations, including different numbers of qubits assigned to the AtomicType.NOUN type. The results showed that a QNLP model can be technically applied to the sentence classification task, but the classification performance remained limited and the overall performance was insufficient for practical extractive summarization. The study identified several major limitations of the method, including the complexity of generated quantum circuits, dataset preparation constraints, simulation stability issues, and limited model generalization.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kontautas, Naglis,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD308337616&prefLang=en_US