Experimental studies on different languages have shown that neurogenetic disorders connected with... more Experimental studies on different languages have shown that neurogenetic disorders connected with Parkinson's disease (PD) determine a series of variations in the speech rhythm. This study aims at verifying whether the speech of PD patients presents rhythmic abnormalities compared to healthy speakers also in Italian. The read speech of 15 healthy speakers and of 11 patients with mild PD was segmented in consonantal and vocalic portions. After extracting the durations of all segments, the vowel percentage (%V) and the interval between two consecutive vowel onset points (VtoV) were calculated. The results show that %V has significantly different values in mildly affected patients as compared to controls. For Italian, %V spans between 44% and 50% for healthy subjects and between 51% and 58% for PD subjects. A positive correlation was found between %V and the number of years of PD since its insurgence. The correlation with the age at which the disease insurges is weak. With regard to VtoV, PD subjects do not speak at a significantly slower rate than healthy controls, though a trend in this direction was found. The data suggest that %V could be used as a more reliable parameter for the early diagnosis of PD than speech rate.
Ihmisarvioijien näkemyksiä suullisen kielitaidon automaattisesta arvioinnista, digitaalisesta arv... more Ihmisarvioijien näkemyksiä suullisen kielitaidon automaattisesta arvioinnista, digitaalisesta arviointiprosessista sekä puhesuorituksista arvioitavista ulottuvuuksista This study investigated human raters' perceptions of automated assessment of oral language skills. The raters (n = 37) participated in three assessment rounds organized by the DigiTala research project using Moodle and Zoom. The raters assessed Finnish and Swedish learners' speech samples using one holistic and five analytical rating scales created in the project. After the assessment, the raters responded to a questionnaire that included Likert-scale and open-ended questions. Numerical responses were analyzed with descriptive statistics, open responses with content analysis. The raters think that automated scoring could support human rating. The assessment rounds were carried out successfully. The selected dimensions proved to be essential parts of the speaking performances. The results will benefit those working on automated assessment and oral language assessment.
Conference of the International Speech Communication Association, 2016
We present a new tool for training neural network language models (NNLMs), scoring sentences, and... more We present a new tool for training neural network language models (NNLMs), scoring sentences, and generating text. The tool has been written using Python library Theano, which allows researcher to easily extend it and tune any aspect of the training process. Regardless of the flexibility, Theano is able to generate extremely fast native code that can utilize a GPU or multiple CPU cores in order to parallelize the heavy numerical computations. The tool has been evaluated in difficult Finnish and English conversational speech recognition tasks, and significant improvement was obtained over our best back-off n-gram models. The results that we obtained in the Finnish task were compared to those from existing RNNLM and RWTHLM toolkits, and found to be as good or better, while training times were an order of magnitude shorter.
The aim of this special issue is to present current research on assessment that supports language... more The aim of this special issue is to present current research on assessment that supports language learning. The aim of all assessment is to stimulate learning although this customarily applies to formative and dynamic assessment. By using multifaceted assessment practices, language teachers can support students' learning in myriad ways. This is also a key aspect of assessment literacy. However, little is known about assessment that supports language learning in the Finnish context. Therefore, this special issue provides an overview of how assessment can be used to enhance language learning in Finland. The publication starts with an introductory article on formative and dynamic assessment.
This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY
This paper addresses the aspect of the meaningfulness of a national assessment in English as a fo... more This paper addresses the aspect of the meaningfulness of a national assessment in English as a foreign language, applying the fairness framework proposed by Kunnan (2018). We compared students’ performance on two receptive language skills of listening and reading on two subsequent national evaluations in Finland, taken by students at the end of compulsory basic education and at the end of general upper secondary education, respectively. The research questions focus on (1) the relationship between students’ performance on the two tests and their gender, language of schooling and parents’ educational level, and (2) the relationship between the students’ receptive language proficiency at the end of basic education and general upper secondary education. The data were analysed using linear regression and quantile regression analyses. The effect of background variables on the proficiency was stronger for low-performing students than for high-performing students. Moreover, students’ profic...
These last years, there has been a regain of interest in unsupervised sub-lexical and lexical uni... more These last years, there has been a regain of interest in unsupervised sub-lexical and lexical unit discovery. Speech segmentation into phone-like units may be a first interesting step for such a task. In this article, we report speech segmentation experiments in Xitsonga, a less-represented language spoken in South Africa. We chose to use convolutional neural networks (CNN) with FBANK static coefficients as input. The models take binary decisions whether a boundary is present or not at each signal sliding frame. We compare the use of a model trained exclusively on Xitsonga data to the use of a bootstrap model trained on a larger corpus of another language, the BUCKEYE U.S. English corpus. Using a two-convolution-layer model, a 79% F-measure was obtained on BUCKEYE, with a 20 ms error tolerance. This performance is equal to the human inter-annotator agreement rate. We then used this bootstrap model to segment Xitsonga data and compared the results when adapting it with 1 to 20 minutes of Xitsonga data.
This qualitative study explores how native peers as teaching assistants can support language lear... more This qualitative study explores how native peers as teaching assistants can support language learning in Higher Education. The data of five semi-structured group interviews of 15 teaching assistants in total were analysed thematically. The results indicate that the assistants embraced multiple roles during the programme and within the lessons. However, in the role of a peer student who collaborated on an equal level with the learners, they could best facilitate learning and communication. As native speakers, they complemented the teacher’s expertise by their regional and up-to-date cultural and linguistic knowledge. This article supports a wide view of mediation adopted by the CEFR Companion Volume and focuses on its two mediation activities. The so-called didactic triangle was used to illustrate how native peers mediated between the teacher, the student and the subject to facilitate the language learning process. The assistants acted as mediators in different ways, interculturally ...
Tämä katsaus tarkastelee kielididaktiikan ja aivotutkimuksen kehitysvaiheita aikaisemman tutkimuk... more Tämä katsaus tarkastelee kielididaktiikan ja aivotutkimuksen kehitysvaiheita aikaisemman tutkimuksen pohjalta sekä osoittaa yhtymäkohtia kielididaktiikan, kognitiivisen psykologian ja aivotutkimuksen välillä. Artikkeli kuvaa sitä, miten käyttäytymisen ja aivojen tutkimus on lisännyt ymmärrystä kielen oppimisen mikrotasosta ja miten tätä tietoa voisi hyödyntää opetuksessa. Lopuksi pohditaan aivotutkimuksen menetelmien hyödyntämisen tulevaisuuden näkymiä sekä niihin mahdollisesti liittyviä riskejä. Aivotutkimuksen ja kielididaktiikan yhteyksien havaitseminen auttaa valaisemaan ihmisen ainutlaatuisinta ominaisuutta: kielen avulla olemme rakentaneet sivilisaation, ja uusia kieliä oppimalla voimme kehittää sitä edelleen.
Experimental studies on different languages have shown that neurogenetic disorders connected with... more Experimental studies on different languages have shown that neurogenetic disorders connected with Parkinson's disease (PD) determine a series of variations in the speech rhythm. This study aims at verifying whether the speech of PD patients presents rhythmic abnormalities compared to healthy speakers also in Italian. The read speech of 15 healthy speakers and of 11 patients with mild PD was segmented in consonantal and vocalic portions. After extracting the durations of all segments, the vowel percentage (%V) and the interval between two consecutive vowel onset points (VtoV) were calculated. The results show that %V has significantly different values in mildly affected patients as compared to controls. For Italian, %V spans between 44% and 50% for healthy subjects and between 51% and 58% for PD subjects. A positive correlation was found between %V and the number of years of PD since its insurgence. The correlation with the age at which the disease insurges is weak. With regard to VtoV, PD subjects do not speak at a significantly slower rate than healthy controls, though a trend in this direction was found. The data suggest that %V could be used as a more reliable parameter for the early diagnosis of PD than speech rate.
Ihmisarvioijien näkemyksiä suullisen kielitaidon automaattisesta arvioinnista, digitaalisesta arv... more Ihmisarvioijien näkemyksiä suullisen kielitaidon automaattisesta arvioinnista, digitaalisesta arviointiprosessista sekä puhesuorituksista arvioitavista ulottuvuuksista This study investigated human raters' perceptions of automated assessment of oral language skills. The raters (n = 37) participated in three assessment rounds organized by the DigiTala research project using Moodle and Zoom. The raters assessed Finnish and Swedish learners' speech samples using one holistic and five analytical rating scales created in the project. After the assessment, the raters responded to a questionnaire that included Likert-scale and open-ended questions. Numerical responses were analyzed with descriptive statistics, open responses with content analysis. The raters think that automated scoring could support human rating. The assessment rounds were carried out successfully. The selected dimensions proved to be essential parts of the speaking performances. The results will benefit those working on automated assessment and oral language assessment.
Conference of the International Speech Communication Association, 2016
We present a new tool for training neural network language models (NNLMs), scoring sentences, and... more We present a new tool for training neural network language models (NNLMs), scoring sentences, and generating text. The tool has been written using Python library Theano, which allows researcher to easily extend it and tune any aspect of the training process. Regardless of the flexibility, Theano is able to generate extremely fast native code that can utilize a GPU or multiple CPU cores in order to parallelize the heavy numerical computations. The tool has been evaluated in difficult Finnish and English conversational speech recognition tasks, and significant improvement was obtained over our best back-off n-gram models. The results that we obtained in the Finnish task were compared to those from existing RNNLM and RWTHLM toolkits, and found to be as good or better, while training times were an order of magnitude shorter.
The aim of this special issue is to present current research on assessment that supports language... more The aim of this special issue is to present current research on assessment that supports language learning. The aim of all assessment is to stimulate learning although this customarily applies to formative and dynamic assessment. By using multifaceted assessment practices, language teachers can support students' learning in myriad ways. This is also a key aspect of assessment literacy. However, little is known about assessment that supports language learning in the Finnish context. Therefore, this special issue provides an overview of how assessment can be used to enhance language learning in Finland. The publication starts with an introductory article on formative and dynamic assessment.
This article is an open access article distributed under the terms and conditions of the Creative... more This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY
This paper addresses the aspect of the meaningfulness of a national assessment in English as a fo... more This paper addresses the aspect of the meaningfulness of a national assessment in English as a foreign language, applying the fairness framework proposed by Kunnan (2018). We compared students’ performance on two receptive language skills of listening and reading on two subsequent national evaluations in Finland, taken by students at the end of compulsory basic education and at the end of general upper secondary education, respectively. The research questions focus on (1) the relationship between students’ performance on the two tests and their gender, language of schooling and parents’ educational level, and (2) the relationship between the students’ receptive language proficiency at the end of basic education and general upper secondary education. The data were analysed using linear regression and quantile regression analyses. The effect of background variables on the proficiency was stronger for low-performing students than for high-performing students. Moreover, students’ profic...
These last years, there has been a regain of interest in unsupervised sub-lexical and lexical uni... more These last years, there has been a regain of interest in unsupervised sub-lexical and lexical unit discovery. Speech segmentation into phone-like units may be a first interesting step for such a task. In this article, we report speech segmentation experiments in Xitsonga, a less-represented language spoken in South Africa. We chose to use convolutional neural networks (CNN) with FBANK static coefficients as input. The models take binary decisions whether a boundary is present or not at each signal sliding frame. We compare the use of a model trained exclusively on Xitsonga data to the use of a bootstrap model trained on a larger corpus of another language, the BUCKEYE U.S. English corpus. Using a two-convolution-layer model, a 79% F-measure was obtained on BUCKEYE, with a 20 ms error tolerance. This performance is equal to the human inter-annotator agreement rate. We then used this bootstrap model to segment Xitsonga data and compared the results when adapting it with 1 to 20 minutes of Xitsonga data.
This qualitative study explores how native peers as teaching assistants can support language lear... more This qualitative study explores how native peers as teaching assistants can support language learning in Higher Education. The data of five semi-structured group interviews of 15 teaching assistants in total were analysed thematically. The results indicate that the assistants embraced multiple roles during the programme and within the lessons. However, in the role of a peer student who collaborated on an equal level with the learners, they could best facilitate learning and communication. As native speakers, they complemented the teacher’s expertise by their regional and up-to-date cultural and linguistic knowledge. This article supports a wide view of mediation adopted by the CEFR Companion Volume and focuses on its two mediation activities. The so-called didactic triangle was used to illustrate how native peers mediated between the teacher, the student and the subject to facilitate the language learning process. The assistants acted as mediators in different ways, interculturally ...
Tämä katsaus tarkastelee kielididaktiikan ja aivotutkimuksen kehitysvaiheita aikaisemman tutkimuk... more Tämä katsaus tarkastelee kielididaktiikan ja aivotutkimuksen kehitysvaiheita aikaisemman tutkimuksen pohjalta sekä osoittaa yhtymäkohtia kielididaktiikan, kognitiivisen psykologian ja aivotutkimuksen välillä. Artikkeli kuvaa sitä, miten käyttäytymisen ja aivojen tutkimus on lisännyt ymmärrystä kielen oppimisen mikrotasosta ja miten tätä tietoa voisi hyödyntää opetuksessa. Lopuksi pohditaan aivotutkimuksen menetelmien hyödyntämisen tulevaisuuden näkymiä sekä niihin mahdollisesti liittyviä riskejä. Aivotutkimuksen ja kielididaktiikan yhteyksien havaitseminen auttaa valaisemaan ihmisen ainutlaatuisinta ominaisuutta: kielen avulla olemme rakentaneet sivilisaation, ja uusia kieliä oppimalla voimme kehittää sitä edelleen.
Uploads
Papers by Raili Hilden