{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/improving-quality-of-hierarchical-clustering","title":"Improving Quality of Hierarchical Clustering for Large Data Series","arxiv_id":"1608.01238","date":"2016-08-03","proceeding":null,"authors":["Manuel R. Ciosici"],"abstract":"Brown clustering is a hard, hierarchical, bottom-up clustering of words in a\nvocabulary. Words are assigned to clusters based on their usage pattern in a\ngiven corpus. The resulting clusters and hierarchical structure can be used in\nconstructing class-based language models and for generating features to be used\nin NLP tasks. Because of its high computational cost, the most-used version of\nBrown clustering is a greedy algorithm that uses a window to restrict its\nsearch space. Like other clustering algorithms, Brown clustering finds a\nsub-optimal, but nonetheless effective, mapping of words to clusters. Because\nof its ability to produce high-quality, human-understandable cluster, Brown\nclustering has seen high uptake the NLP research community where it is used in\nthe preprocessing and feature generation steps.\n  Little research has been done towards improving the quality of Brown\nclusters, despite the greedy and heuristic nature of the algorithm. The\napproaches tried so far have focused on: studying the effect of the\ninitialisation in a similar algorithm; tuning the parameters used to define the\ndesired number of clusters and the behaviour of the algorithm; and including a\nseparate parameter to differentiate the window from the desired number of\nclusters. However, some of these approaches have not yielded significant\nimprovements in cluster quality.\n  In this thesis, a close analysis of the Brown algorithm is provided,\nrevealing important under-specifications and weaknesses in the original\nalgorithm. These have serious effects on cluster quality and reproducibility of\nresearch using Brown clustering. In the second part of the thesis, two\nmodifications are proposed. Finally, a thorough evaluation is performed,\nconsidering both the optimization criterion of Brown clustering and the\nperformance of the resulting class-based language models.","url_abs":"http://arxiv.org/abs/1608.01238v1","url_pdf":"http://arxiv.org/pdf/1608.01238v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"improving-quality-of-hierarchical-clustering","repo_url":"https://github.com/manuelciosici/ExchangeAndBrown","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"clustering","task_name":"Clustering"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}