Elamur Research⌕
☰
home / papers / 2026 / Efficient Topic Extraction via Graph-Based Labeling: A Lightweight Alternative to Deep Models

Efficient Topic Extraction via Graph-Based Labeling: A Lightweight Alternative to Deep Models

Summary

Topic models such as LDA describe each topic only as a ranked list of words, which readers often struggle to interpret. Recent labeling solutions based on large neural models work better but are expensive to run. Mekaoui et al. propose a lightweight and transparent alternative for topic labeling. Their method treats a topic's keywords as one concatenated text unit rather than as independent terms. It then picks a label with a single similarity procedure instead of combining many graph metrics.

The authors test two variants. DSL works on the original keyword set. GEL first expands the keywords with semantically related concepts from the ConceptNet knowledge graph. Both variants are run with several sentence-embedding models and evaluated on Topic_Bhatia and 20 Newsgroups. The comparison covers earlier labelers, a fine-tuned BART-TL, and ChatGPT-3.5.

The proposed methods beat prior baselines on BERTScore and come close to ChatGPT-3.5 on cosine similarity while using far less compute. This suggests that simple graph-based labeling is a practical option when resources are limited or explainability matters.

Key points

  • BERTScore results: DSL (all-MiniLM-L6-v2) and GEL (gtr-t5-large) both reach a BERTScore F1 of 0.955. The best earlier baseline (Top2Label_all_relations) scores 0.936, and the BART-TL variants score 0.886–0.892.
  • Cosine similarity on 20 Newsgroups: GEL with GIST-all-MiniLM-L6-v2 reaches 0.627 and DSL with all-MiniLM-L12-v2 reaches 0.578. ChatGPT-3.5 scores 0.655 and BART-TL about 0.31, so GEL trails ChatGPT-3.5 by only 0.028.
  • Effect of graph enrichment: Expanding the keywords through ConceptNet (GEL) generally gives higher cosine scores than using the raw keywords alone (DSL).
  • Embedding models tested: all-MiniLM-L6-v2, all-MiniLM-L12-v2, all-mpnet-base-v2, gtr-t5-large, and the GIST family (small, large, all-MiniLM-L6-v2).
  • Limitations and future work:
    • Topic_Bhatia allows several valid labels per topic, which makes a consistent gold standard hard to define.
    • 20 Newsgroups offers only 20 labels.
    • The authors call for larger dedicated labeling datasets, other knowledge graphs, and techniques such as graph2vec.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/5q74w4fjj6sw

@inproceedings{mekaoui2026efficient,
  title = {{Efficient Topic Extraction via Graph-Based Labeling: A Lightweight Alternative to Deep Models}},
  author = {Mekaoui, Salma and Sofyan, Hiba and Benchrif, Imane and Amaaz, Imane and Chaker, Ilham and Zarghili, Arsalane and Nikolov, Nikola S.},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/5q74w4fjj6sw},
  url = {https://doi.org/10.63317/5q74w4fjj6sw}
}

References

57 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Rifki Adhitama, Retno Kusumaningrum, and Rahmat Gernowo. 2017. Topic labeling towards news document collection based on latent dirichlet allocation and ontology. In 2017 1st International Conference on Informatics and Computational Sciences (ICICoS), pages 247–252. IEEE. find
  2. Mehdi Allahyari and Krys Kochut. 2015. Automatic topic labeling using ontology-based topic models. In 2015 IEEE 14th International Conference on Machine Learning and Applications (ICMLA), pages 259–264. IEEE. find
  3. Areej Alokaili, Nikolaos Aletras, and Mark Stevenson. 2020. Automatic generation of topic labels. In Proceedings of the 43rd international ACM SI-GIR conference on research and development in information retrieval, pages 1965–1968. find
  4. Shraey Bhatia, Jey Han Lau, and Timothy Baldwin. 2016. Automatic labelling of topics with neural embeddings. arXiv preprint arXiv:1612.05340. find
  5. Ehsan Bitaraf, Maryam Jafarpour, Sina Shool, and Reza Saboori Amleshi. 2024. Unveiling medical insights: Advanced topic extraction from scientific articles. In Digital Health and Informatics Innovations for Sustainable Health Care Systems, pages 944–948. IOS Press. find
  6. Akhil Chaudhary, Evangelos Milios, and Enayat Rajabi. 2024. Top2label: Explainable zero shot topic labelling using knowledge graphs. Expert Systems with Applications, 242:122676. find
  7. Kenneth Ward Church. 2017. Word2vec. Natural Language Engineering, 23(1):155–162. find
  8. Rob Churchill and Lisa Singh. 2022. The evolution of topic modeling. ACM Computing Surveys, 54(10s):1–35. find
  9. Heidar Davoudi and Aijun An. 2015. Ontology-based topic labeling and quality prediction. In International Symposium on Methodologies for Intelligent Systems, pages 171–179. Springer. find
  10. Fellyca Effendi, Muhammad Rizky Pribadi, Wijang Widhiarso, and Sherly Devyanes. 2023. Hybrid model for topic classification of english learning accounts on instagram using lda and svm. In 2023 10th International Conference on Electrical Engineering, Computer Science and Informatics (EECSI), pages 64… find
  11. Elena Flondor and Marc Frincu. 2023. Analysing android apps classification and categories validation by using latent dirichlet allocation. In International Conference on Computational Collective Intelligence, pages 282–297. Springer. find
  12. Antoine Gourru, Julien Velcin, Mathieu Roche, Christophe Gravier, and Pascal Poncelet. 2018. United we stand: Using multiple strategies for topic labeling. In International Conference on Applications of Natural Language to Information Systems, pages 352–363. Springer. find
  13. Gerhard Johann Hagerer, Martin Kirchhoff, Hannah Danner, Robert Pesch, Mainak Ghosh, Archishman Roy, Jiaxi Zhao, and Georg Groh. 2021. Socialvistum: An interactive visualization toolkit for correlated neural topic models on social media opinion mining. arXiv preprint arXiv:2110.10575. find
  14. Dongbin He, Minjuan Wang, Abdul Mateen Khattak, Li Zhang, and Wanlin Gao. 2019. Automatic labeling of topic models using graph-based ranking. IEEE Access, 7:131593–131608. find
  15. Syeda Sumbul Hossain, Md Rezwan Ul-Hassan, and Shadikur Rahman. 2019. Polynomial topic distribution with topic modeling for generic labeling. In International Conference on Advances in Computing and Data Sciences, pages 409–419. Springer. find
  16. Hugging Face. 2025. Massive text embedding benchmark (mteb) leaderboard. https://huggingface.co/mteb. Accessed: 2025-09-16. ↗
  17. Hyon Hee Kim and Hey Young Rhee. 2019. An ontology-based labeling of influential topics using topic network analysis. Journal of Information Processing Systems, 15(5):1096–1107. find
  18. Supriya Ashish Kinariwala and Sachin Deshmukh. 2021. Onto_tml: Auto-labeling of topic models. Journal of Integrated Science and Technology, 9(2):85–91. find
  19. Damir Korenčić, Strahil Ristov, and Jan Šnajder. 2015. Getting the agenda right: measuring media agenda using topic models. In Proceedings of the 2015 Workshop on Topic Models: Post-Processing and Applications, pages 61–66. find
  20. Olzhas Kozbagarov, Rustam Mussabayev, and Nenad Mladenovic. 2021. A new sentence-based interpretative topic modeling and automatic topic labeling. Symmetry, 13(5):837. find
  21. Jey Han Lau, Karl Grieser, David Newman, and Timothy Baldwin. 2011. Automatic labelling of topic models. In Proceedings of the 49th annual meeting of the association for computational linguistics: human language technologies, pages 1536–1545. find
  22. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov, and Luke Zettlemoyer. 2019. Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. arXiv preprint arXiv:1910.13461. find our card
  23. Chenliang Li, Shiqian Chen, and Yan Qi. 2019. Filtering and classifying relevant short text with a few seed words. Data and Information Management, 3(3):165–186. find
  24. Chenliang Li, Jian Xing, Aixin Sun, and Zongyang Ma. 2016. Effective document labeling with very few seed words: A topic model approach. In Proceedings of the 25th ACM international on conference on information and knowledge management, pages 85–94. find
  25. Pengyu Li, Christine Tseng, Yaxuan Zheng, Joyce A Chew, Longxiu Huang, Benjamin Jarman, and Deanna Needell. 2022. Guided semi-supervised non-negative matrix factorization on legal documents. arXiv preprint arXiv:2201.13324. — [arXiv:2201.13324](https://arxiv.org/abs/2201.13324) find
  26. Xiaofei Li and Guo Chen. Automatic literature review generation by integrating large and small models. find
  27. Hugo Liu and Push Singh. 2004. Conceptnet—a practical commonsense reasoning tool-kit. BT technology journal, 22(4):211–226. find
  28. Lin Liu and Lin Tang. 2019. A label distribution topic model for multi-label classification. In Proceedings of the 4th International Conference on Intelligent Information Processing, pages 52–57. find
  29. Davide Magatti, Silvia Calegari, Davide Ciucci, and Fabio Stella. 2009. Automatic labeling of topics. In 2009 Ninth International Conference on Intelligent Systems Design and Applications, pages 1227–1232. IEEE. find
  30. Xian-Ling Mao, Zhao-Yan Ming, Zheng-Jun Zha, Tat-Seng Chua, Hongfei Yan, and Xiaoming Li. 2012. Automatic labeling hierarchical topics. In Proceedings of the 21st ACM international conference on Information and knowledge management, pages 2383–2386. find
  31. Xian-Ling Mao, Yixuan Xiao, Qiang Zhou, Jun Wang, and Heyan Huang. 2015. Ehllda: a supervised hierarchical topic model. In China National Conference on Chinese Computational Linguistics, pages 215–226. Springer. find
  32. Xian-Ling Mao, Yi-jing Zhao, Qiang Zhou, Wen-Qing Yuan, Liner Yang, and He-Yan Huang. 2016. A novel fast framework for topic labeling based on similarity-preserved hashing. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3339–334… find
  33. Salma Mekaoui, Ilham Chaker, Arsalane Zarghili, and Nikola S Nikolov. 2025. Systematic literature review of topic labeling. IEEE Access. find
  34. Tomas Mikolov, Ilya Sutskever, Kai Chen, Greg S Corrado, and Jeff Dean. 2013. Distributed representations of words and phrases and their compositionality. Advances in neural information processing systems, 26. find
  35. Yida Mu, Chun Dong, Kalina Bontcheva, and Xingyi Song. 2024. Large language models offer an alternative to the traditional approach of topic modelling. arXiv preprint arXiv:2403.16248. find
  36. Takafumi Nakanishi. 2025. Fusion of xai and llm: Automatic topic name extraction from text data using approximate inverse model explanations (aime). In 2025 19th International Conference on Semantic Computing (ICSC), pages 203–208. IEEE Computer Society. find
  37. Jianmo Ni, Chen Qu, Jing Lu, Zhuyun Dai, Gustavo Hernandez Abrego, Ji Ma, Vincent Y Zhao, Yi Luan, Keith B Hall, Ming-Wei Chang, et al. 2021. Large dual encoders are generalizable retrievers. arXiv preprint arXiv:2112.07899. find our card
  38. Diogo Nolasco and Jonice Oliveira. 2016. Detecting knowledge innovation through automatic topic labeling on scholar data. In 2016 49th Hawaii International Conference on System Sciences (HICSS), pages 358–367. IEEE. find
  39. Takuro Ogawa and Ryosuke Saga. 2021. Text-based causality modeling with a conceptual label in a hierarchical topic structure using bayesian rose trees. find
  40. OpenAI. 2023. Introducing chatgpt. https://openai.com/blog/chatgpt. Accessed: 2025-10-07. ↗
  41. Cristian Popa and Traian Rebedea. 2021. Bart-tl: Weakly-supervised topic label generation. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 1418–1425. find
  42. Virginia Ramón-Ferrer, Carlos Badenes-Olmedo, and Oscar Corcho. 2023. Automatic topic label generation using conversational models. In Proceedings of the 12th Knowledge Capture Conference 2023, pages 17–24. find
  43. Michael Reynolds. 2025. Research on the application of large language models in classification and indexing. find
  44. Ngurah Agus Sanjaya, Mouhamadou Lamine Ba, Talel Abdessalem, and Stéphane Bressan. 2018. Harnessing truth discovery algorithms on the topic labelling problem. In Proceedings of the 20th International Conference on Information Integration and Web-based Applications & Services, pages 8–14. find
  45. Thomas Scelsi, Alfonso Martinez Arranz, and Lea Frermann. 2021. Principled analysis of energy discourse across domains with thesaurus-based automatic topic labeling. In Proceedings of the 19th Annual Workshop of the Australasian Language Technology Association, pages 107–118. find
  46. Kristian Slabbekoorn, Tomoya Noro, and Takehiro Tokuda. 2016. Ontology-assisted discovery of hierarchical topic clusters on the social web. Journal of Web Engineering, pages 361–396. find
  47. Aivin V. Solatorio. 2024. Gistembed: Guided in-sample selection of training negatives for text embedding fine-tuning. arXiv preprint arXiv:2402.16829. — [arXiv:2402.16829](https://arxiv.org/abs/2402.16829) find
  48. Wenling Tang, Xu Wu, Yuxiao Li, and Jin Xu. 2016. A topic label extraction method for the university bbs. In 2016 IEEE First International Conference on Data Science in Cyberspace (DSC), pages 678–682. IEEE. find
  49. Paras Tiwari, Ashutosh Tripathi, Avaneesh Singh, and Sawan Rai. 2023a. Advanced hierarchical topic labeling for short text. IEEE access, 11:35158–35174. find
  50. Paras Tiwari, Ashutosh Tripathi, Avaneesh Singh, and Sawan Rai. 2023b. Advanced hierarchical topic labeling for short text. IEEE access, 11:35158–35174. find
  51. Eleni Vathi, Georgios Siolas, and Andreas Stafylopatis. 2017. Mining and categorizing interesting topics in twitter communities. Journal of Intelligent & Fuzzy Systems, 32(2):1265–1275. find
  52. Yiming Wang, Jindong Zhang, Zhiyao Yang, Bing Wang, Jingyi Jin, and Yitong Liu. 2024. Improving extractive summarization with semantic enhancement through topic-injection based bert model. Information Processing & Management, 61(3):103677. find
  53. Justin Wood, Patrick Tan, Wei Wang, and Corey Arnold. 2017. Source-lda: Enhancing probabilistic topic models using prior knowledge sources. In 2017 IEEE 33rd international conference on data engineering (ICDE), pages 411–422. IEEE. find
  54. Daochen Zha and Chenliang Li. 2019. Multi-label dataless text classification with topic modeling. Knowledge and Information Systems, 61(1):137– 160. find
  55. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. 2019. Bertscore: Evaluating text generation with bert. arXiv preprint arXiv:1904.09675. find
  56. Bing Zheng and Hua Li. 2022. Automatic labeling of sdn controller defect text based on neural topic model. In 2022 IEEE 22nd International Conference on Software Quality, Reliability, and Security Companion (QRS-C), pages 402–410. IEEE. find
  57. Elaine Zosa, Lidia Pivovarova, Michele Boggia, and Sardana Ivanova. 2022. Multilingual topic labelling of news topics using ontological mapping. In European Conference on Information Retrieval, pages 248–256. Springer. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/papers/mekaoui-2026-efficient-topic-extraction-graph-based/