{"id":{"repo_id":"reykjavik","oai_identifier":"oai:skemman.is:1946/50888"},"canonical_url":"https://search.dev.ndltd.org/etd/reykjavik/oai:skemman.is:1946/50888","repository":{"repo_id":"reykjavik","name":"Reykjavík University","base_url":"https://skemman.is/oai/request"},"display":{"title":"Analyzing Icelandic conversation using State-of-the-Art ASR models","abstract":"This thesis investigates the performance and fine-tuning of transformer-based Automatic Speech Recognition (ASR) systems applied to Icelandic conversational speech, with a primary focus on OpenAI’s Whisper model and a secondary fo- cus on Meta’s Wav2Vec 2.0. The study uses the Spjallrómur dataset, a 21-hour unscripted dialogue dataset between two speakers for benchmarking and model training. The project examines both reduced parameter number Low-Rank Adap- tation ( LoRA) and full parameter fine-tuning strategies. In this thesis, only full- parameter fine-tuning was implemented and tested. Both the base Whisper-Small and Whisper-Large models are used, starting from a model pre-trained on 967 hours of read Icelandic speech, which served as the basis for further fine-tuning. Fine-tuning from the 967h pre-trained Whisper-Large model reduced the WER from 29.84% to 22.69% and the CER from 20.48% to 13.10%. This indicates that further fine-tuning with low-quality data yields a notable improvement. Whisper- small improved considerably, reducing its Word Error Rate (WER) from 145.71% to WER of 48.30% (with CER of 29.37%). These findings indicate that even a small and misaligned dataset can yield substantial improvement on both OpenAI’s - Small and Whisper-Large. Keywords/Efnisorð: Whisper, Wav2Vec 2.0, K2, Icelandic, ASR, fine-tuning, LoRA, transformer models, conversational speech","abstract_html":"This thesis investigates the performance and fine-tuning of transformer-based Automatic Speech Recognition (ASR) systems applied to Icelandic conversational speech, with a primary focus on OpenAI’s Whisper model and a secondary fo- cus on Meta’s Wav2Vec 2.0. The study uses the Spjallrómur dataset, a 21-hour unscripted dialogue dataset between two speakers for benchmarking and model training. The project examines both reduced parameter number Low-Rank Adap- tation ( LoRA) and full parameter fine-tuning strategies. In this thesis, only full- parameter fine-tuning was implemented and tested. Both the base Whisper-Small and Whisper-Large models are used, starting from a model pre-trained on 967 hours of read Icelandic speech, which served as the basis for further fine-tuning. Fine-tuning from the 967h pre-trained Whisper-Large model reduced the WER from 29.84% to 22.69% and the CER from 20.48% to 13.10%. This indicates that further fine-tuning with low-quality data yields a notable improvement. Whisper- small improved considerably, reducing its Word Error Rate (WER) from 145.71% to WER of 48.30% (with CER of 29.37%). These findings indicate that even a small and misaligned dataset can yield substantial improvement on both OpenAI’s - Small and Whisper-Large. Keywords/Efnisorð: Whisper, Wav2Vec 2.0, K2, Icelandic, ASR, fine-tuning, LoRA, transformer models, conversational speech","abstract_has_math":false,"creators":["Páll Rúnarsson 1982-"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":["Háskólinn í Reykjavík"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2025,"date_issued":"2025-06-18T14:39:16Z","date_published":"2025-06-18T14:39:16Z","updated_at":"2026-07-27T20:38:22Z","subjects":["Meistaraprófsritgerðir","Verkfræði","Talkennsl","Máltækni","Gervigreind","Engineering","Speech recognition","Transformer model","Conversational speech"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/1946/50888","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Háskólinn í Reykjavík"]},{"key":"dc:creator","label":"Author","values":["Páll Rúnarsson 1982-"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-06-18T14:39:15Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-06-18T14:39:15Z"]},{"key":"dc:date.issued","label":"Date","values":["2025-06-18T14:39:16Z"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Meistaraprófsritgerðir","Verkfræði","Talkennsl","Máltækni","Gervigreind","Engineering","Speech recognition","Transformer model","Conversational speech"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1946/50888"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This thesis investigates the performance and fine-tuning of transformer-based Automatic Speech Recognition (ASR) systems applied to Icelandic conversational speech, with a primary focus on OpenAI’s Whisper model and a secondary fo- cus on Meta’s Wav2Vec 2.0. The study uses the Spjallrómur dataset, a 21-hour unscripted dialogue dataset between two speakers for benchmarking and model training. The project examines both reduced parameter number Low-Rank Adap- tation ( LoRA) and full parameter fine-tuning strategies. In this thesis, only full- parameter fine-tuning was implemented and tested. Both the base Whisper-Small and Whisper-Large models are used, starting from a model pre-trained on 967 hours of read Icelandic speech, which served as the basis for further fine-tuning. Fine-tuning from the 967h pre-trained Whisper-Large model reduced the WER from 29.84% to 22.69% and the CER from 20.48% to 13.10%. This indicates that further fine-tuning with low-quality data yields a notable improvement. Whisper- small improved considerably, reducing its Word Error Rate (WER) from 145.71% to WER of 48.30% (with CER of 29.37%). These findings indicate that even a small and misaligned dataset can yield substantial improvement on both OpenAI’s - Small and Whisper-Large. Keywords/Efnisorð: Whisper, Wav2Vec 2.0, K2, Icelandic, ASR, fine-tuning, LoRA, transformer models, conversational speech"]},{"key":"dc:title","label":"Title","values":["Analyzing Icelandic conversation using State-of-the-Art ASR models"]}]}],"canonical_facts":{"dc:contributor":["Háskólinn í Reykjavík"],"dc:creator":["Páll Rúnarsson 1982-"],"dc:date.accessioned":["2025-06-18T14:39:15Z"],"dc:date.available":["2025-06-18T14:39:15Z"],"dc:date.issued":["2025-06-18T14:39:16Z"],"dc:description.abstract":["This thesis investigates the performance and fine-tuning of transformer-based Automatic Speech Recognition (ASR) systems applied to Icelandic conversational speech, with a primary focus on OpenAI’s Whisper model and a secondary fo- cus on Meta’s Wav2Vec 2.0. The study uses the Spjallrómur dataset, a 21-hour unscripted dialogue dataset between two speakers for benchmarking and model training. The project examines both reduced parameter number Low-Rank Adap- tation ( LoRA) and full parameter fine-tuning strategies. In this thesis, only full- parameter fine-tuning was implemented and tested. Both the base Whisper-Small and Whisper-Large models are used, starting from a model pre-trained on 967 hours of read Icelandic speech, which served as the basis for further fine-tuning. Fine-tuning from the 967h pre-trained Whisper-Large model reduced the WER from 29.84% to 22.69% and the CER from 20.48% to 13.10%. This indicates that further fine-tuning with low-quality data yields a notable improvement. Whisper- small improved considerably, reducing its Word Error Rate (WER) from 145.71% to WER of 48.30% (with CER of 29.37%). These findings indicate that even a small and misaligned dataset can yield substantial improvement on both OpenAI’s - Small and Whisper-Large. Keywords/Efnisorð: Whisper, Wav2Vec 2.0, K2, Icelandic, ASR, fine-tuning, LoRA, transformer models, conversational speech"],"dc:identifier.uri":["https://hdl.handle.net/1946/50888"],"dc:language.iso":["en"],"dc:subject":["Meistaraprófsritgerðir","Verkfræði","Talkennsl","Máltækni","Gervigreind","Engineering","Speech recognition","Transformer model","Conversational speech"],"dc:title":["Analyzing Icelandic conversation using State-of-the-Art ASR models"],"dc:type":["Thesis"]},"updated_at":"2026-07-27T20:38:22Z"}