{"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/truly-unsupervised-acoustic-word-embeddings","title":"Truly unsupervised acoustic word embeddings using weak top-down constraints in encoder-decoder models","arxiv_id":"1811.00403","date":"2018-11-01","proceeding":null,"authors":["Herman Kamper"],"abstract":"We investigate unsupervised models that can map a variable-duration speech\nsegment to a fixed-dimensional representation. In settings where unlabelled\nspeech is the only available resource, such acoustic word embeddings can form\nthe basis for \"zero-resource\" speech search, discovery and indexing systems.\nMost existing unsupervised embedding methods still use some supervision, such\nas word or phoneme boundaries. Here we propose the encoder-decoder\ncorrespondence autoencoder (EncDec-CAE), which, instead of true word segments,\nuses automatically discovered segments: an unsupervised term discovery system\nfinds pairs of words of the same unknown type, and the EncDec-CAE is trained to\nreconstruct one word given the other as input. We compare it to a standard\nencoder-decoder autoencoder (AE), a variational AE with a prior over its latent\nembedding, and downsampling. EncDec-CAE outperforms its closest competitor by\n24% relative in average precision on two languages in a word discrimination\ntask.","url_abs":"http://arxiv.org/abs/1811.00403v2","url_pdf":"http://arxiv.org/pdf/1811.00403v2.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":"truly-unsupervised-acoustic-word-embeddings","repo_url":"https://github.com/kamperh/recipe_bucktsong_awe","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"truly-unsupervised-acoustic-word-embeddings","repo_url":"https://github.com/kamperh/recipe_bucktsong_awe_py3","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"word-embeddings","task_name":"Word Embeddings"}],"methods":[{"method_slug":"ae","method_name":"AE"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1811.00403","atlas_url":"https://app.syntology.ai/?focus=1811.00403","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}