TY - GEN

T1 - Key Algorithms for Keyphrase Generation: Instruction-Based LLMs for Russian Scientific Keyphrases

AU - Glazkova, Anna

AU - Morozov, Dmitry

AU - Garipov, Timur

N1 - Key Algorithms for Keyphrase Generation: Instruction-Based LLMs for Russian Scientific Keyphrases / A. Glazkova, D. Morozov, T. Garipov // Analysis of Images, Social Networks and Texts. AIST 2024. Lecture Notes in Computer Science. / A. Panchenko [et al] // Springer, Cham.: 2025. - vol 15419. DOI: https://doi.org/10.1007/978-3-031-88036-0_5

PY - 2025/4/15

Y1 - 2025/4/15

N2 - Keyphrase selection is a challenging task in natural language processing that has a wide range of applications. Adapting existing supervised and unsupervised solutions for the Russian language faces several limitations due to the rich morphology of Russian and the limited number of training datasets available. Recent studies conducted on English texts show that large language models (LLMs) successfully address the task of generating keyphrases. LLMs allow achieving impressive results without task-specific fine-tuning, using text prompts instead. In this work, we access the performance of prompt-based methods for generating keyphrases for Russian scientific abstracts. First, we compare the performance of zero-shot and few-shot prompt-based methods, fine-tuned models, and unsupervised methods. Then we assess strategies for selecting keyphrase examples in a few-shot setting. We present the outcomes of human evaluation of the generated keyphrases and analyze the strengths and weaknesses of the models through expert assessment. Our results suggest that prompt-based methods can outperform common baselines even using simple text prompts.

AB - Keyphrase selection is a challenging task in natural language processing that has a wide range of applications. Adapting existing supervised and unsupervised solutions for the Russian language faces several limitations due to the rich morphology of Russian and the limited number of training datasets available. Recent studies conducted on English texts show that large language models (LLMs) successfully address the task of generating keyphrases. LLMs allow achieving impressive results without task-specific fine-tuning, using text prompts instead. In this work, we access the performance of prompt-based methods for generating keyphrases for Russian scientific abstracts. First, we compare the performance of zero-shot and few-shot prompt-based methods, fine-tuned models, and unsupervised methods. Then we assess strategies for selecting keyphrase examples in a few-shot setting. We present the outcomes of human evaluation of the generated keyphrases and analyze the strengths and weaknesses of the models through expert assessment. Our results suggest that prompt-based methods can outperform common baselines even using simple text prompts.

KW - Keyphrase selection

KW - Large language model

KW - Prompt-based learning

KW - Scientific texts

UR - https://www.mendeley.com/catalogue/150db0f7-cc37-30f1-9ae7-6b0af5cf02db/

UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-105003859167&origin=inward&txGid=e6a897f76c6f2da113c1d1a0081684cf

U2 - 10.1007/978-3-031-88036-0_5

DO - 10.1007/978-3-031-88036-0_5

M3 - Conference contribution

SN - 9783031880353

T3 - Lecture Notes in Computer Science

SP - 107

EP - 119

BT - Lecture Notes in Computer Science

PB - Springer

ER -