Lahjoita puhetta: a large-scale corpus of spoken Finnish with some benchmarks

Moisio, Anssi; Porjazovski, Dejan; Rouhe, Aku; Getman, Yaroslav; Virkkunen, Anja; Al-Ghezi, Ragheb; Lennes, Mietta; Grósz, Tamás; Lindén, Krister; Kurimo, Mikko

doi:10.1007/s10579-022-09606-3

Cited by 12 publications

(12 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DSPCon (Enarvi, 2018 ), FinDialogue (Lennes, 2009 ), and Lahjoita puhetta (Moisio et al, 2022 ) corpora represent more spontaneous and conversational forms of Finnish speech. DSPCon contains short conversations between students.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Finnish parliament ASR corpus

Virkkunen

Rouhe

Phan

et al. 2023

Lang Resources & Evaluation

Self Cite

View full text Add to dashboard Cite

Public sources like parliament meeting recordings and transcripts provide ever-growing material for the training and evaluation of automatic speech recognition (ASR) systems. In this paper, we publish and analyse the Finnish Parliament ASR Corpus, the most extensive publicly available collection of manually transcribed speech data for Finnish with over 3000 h of speech and 449 speakers for which it provides rich demographic metadata. This corpus builds on earlier initial work, and as a result the corpus has a natural split into two training subsets from two periods of time. Similarly, there are two official, corrected test sets covering different times, setting an ASR task with longitudinal distribution-shift characteristics. An official development set is also provided. We developed a complete Kaldi-based data preparation pipeline and ASR recipes for hidden Markov models (HMM), hybrid deep neural networks (HMM-DNN), and attention-based encoder-decoders (AED). For HMM-DNN systems, we provide results with time-delay neural networks (TDNN) as well as state-of-the-art wav2vec 2.0 pretrained acoustic models. We set benchmarks on the official test sets and multiple other recently used test sets. Both temporal corpus subsets are already large, and we observe that beyond their scale, HMM-TDNN ASR performance on the official test sets has reached a plateau. In contrast, other domains and larger wav2vec 2.0 models benefit from added data. The HMM-DNN and AED approaches are compared in a carefully matched equal data setting, with the HMM-DNN system consistently performing better. Finally, the variation of the ASR accuracy is compared between the speaker categories available in the parliament metadata to detect potential biases based on factors such as gender, age, and education.

show abstract

Section: Related Workmentioning

confidence: 99%

“…This is a known issue with AED models on the Lahjoita puhetta test set. To alleviate the problem, we applied a simple post-processing filter where repetitions are allowed to produce a maximum of five tokens, as originally explained by Moisio et al ( 2022 ).…”

Section: Modelsmentioning

confidence: 99%

Finnish parliament ASR corpus

Virkkunen

Rouhe

Phan

et al. 2023

Lang Resources & Evaluation

Self Cite

View full text Add to dashboard Cite

show abstract

“…To determine whether CL indeed helps the training process of e2e ASR systems, we tested a range of different setups. All of our results are based on the Lahjoita puhetta (Donate speech) corpus [17], which is a collection of colloquial Finnish speech. The dataset consists of over twenty thousand speakers from different regions of Finland and a variety of age groups [17].…”

Section: Datamentioning

confidence: 99%

“…All of our results are based on the Lahjoita puhetta (Donate speech) corpus [17], which is a collection of colloquial Finnish speech. The dataset consists of over twenty thousand speakers from different regions of Finland and a variety of age groups [17]. Recording was performed by the users' own devices, resulting in a range of different noise levels in the data.…”

Section: Datamentioning

confidence: 99%

“…This would result in excessive padding and would prevent us from using a large batch size. To make the data suitable for e2e models, we segmented all long utterances into 10-second chunks, using the best hybrid ASR system described in [17] to detect the silences where the cuts can be performed. This process removed some unnecessary noise and silences, which decreased the final duration of the training set to 1382 hours, the validation set to 8.3 hours and the test set to 8.87 hours.…”

Section: Datamentioning

confidence: 99%

See 1 more Smart Citation

Comparison and Analysis of New Curriculum Criteria for End-to-End ASR

Karakasidis¹,

Grósz²,

Kurimo³

2022

Interspeech 2022

Self Cite

View full text Add to dashboard Cite

It is common knowledge that the quantity and quality of the training data play a significant role in the creation of a good machine learning model. In this paper, we take it one step further and demonstrate that the way the training examples are arranged is also of crucial importance. Curriculum Learning is built on the observation that organized and structured assimilation of knowledge has the ability to enable faster training and better comprehension. When humans learn to speak, they first try to utter basic phones and then gradually move towards more complex structures such as words and sentences. This methodology is known as Curriculum Learning, and we employ it in the context of Automatic Speech Recognition. We hypothesize that end-to-end models can achieve better performance when provided with an organized training set consisting of examples that exhibit an increasing level of difficulty (i.e. a curriculum). To impose structure on the training set and to define the notion of an easy example, we explored multiple scoring functions that either use feedback from an external neural network or incorporate feedback from the model itself. Empirical results show that with different curriculums we can balance the training times and the network's performance.

show abstract