CALST

CALST

The Computer-Assisted Listening and Speaking Tutor (CALST) is a pronunciation training platform which currently offers exercises for English and for several Norwegian dialects. The reason for offering pronunciation training in several dialects is that Norwegian does not have an accepted pronunciation standard and its speakers will use their own dialect independent of the communicative setting.

This means that learners of Norwegian must be able to understand the different dialects, while they need to learn to speak only one of them. Most learners choose the dialect as spoken in Oslo as their target dialect; this dialect has the most extensive training material in CALST.

CALST will also help you to acquire a basic vocabulary of 1000 words and expressions. There are many vocabulary training programs out there, but CALST also offers specific pronunciation exercises, based on sound phonetic knowledge.

Take me to CALST

In the EU Horizon2020 project easyRights, CALST is being extended with the languages of southern European countries which take up many migrants: Greek, Spanish, and (soon) Italian. These languages can already be accessed, but they are still under development.

Take me to testCALST

  • Is pronunciation important?

Absolutely! We most often use a new language in conversations with others. A good pronunciation will make it easier for your conversation partner to understand you, and will help you make a positive first impression, be it in social settings or in a job interview.

  • Pronunciation is a skill

Are you learning a new language, and do you want to work more on your pronunciation? Language courses often pay little attention to pronunciation, because learning the motor skills for new articulations requires repetition. Practice, practice, practice is best done outside of the classroom – as long as you remember to use what you have learnt in real conversations.

  • One size fits all?

Nope. Learners with different native languages meet different challenges. A German learner of English may find it hard to pronounce the first sound in we, while a Chinese learner may find it hard to say zoo instead of Sue. CALST takes your native language into consideration to tailor the exercises specially for you, so that you do not have to waste time and can focus on what may be difficult for you.

  • One for all and all for one

There are still many unsolved questions in language learning. That is why your exercise results are logged and stored anonymously. You thus help to further tailor the exercises for other learners who have the same native language as you. And if a sound is not a problem in one language, you will not get exercises for it in another language you learn later. Be a part of the language learning community, and help to solve the puzzle of language learning.

WELCOME TO THE CALST COMMUNITY!

 

Frequently asked questions

Frequently asked questions

CALST can be run from any pc, tablet or smartphone, irrespective of the operating system. CALST uses a responsive design, which means it automatically adapts to the screen size of your device.

Note that the browser is not allowed to access the microphone on an iPhone or iPad. On these devices, you will therefore only get listening and writing exercises, but no pronunciation exercises. We advise users to take these exercises on any other device (including an Apple laptop).

Yes, CALST is a web-based pronunciation training platform, so you need access to the internet to run CALST – but we do use local storage to deal with short interruptions of your internet connectivity. You can run CALST from most browsers, but it has been fully tested only with Chrome. You can download Chrome for free to your device.
Registered users can see the results from all the exercises they have taken. Also, CALST will remember the last exercise and automatically direct you to the next one.
By registering you allow your results to be logged, so that they can be used to improve CALST and to tailor the exercises for other learners with the same native language.
  1. The pronunciation learning typically starts with two listening exercises. In the first exercise you hear two very similar words, followed by a repetition of one of these words. By comparing the last word to the first two words in your acoustic memory and choosing one, you learn what distinguishes them. This is called a discrimination task.
  2. After the discrimination exercise you take another listening exercise in which you just hear one word. Your task is to say which of two very similar words it is. Since you cannot compare with any words you have just heard, this requires that you have a clear idea of their pronunciation (a so-called internal representation). This task is called an identification task.
  3. Once you have learned to hear the sounds (or other properties, like word stress) correctly, you are ready to pronounce them. By listening to a recording of your own voice and comparing it to the tutor’s, you can adapt and improve your pronunciation.
  4. Finally, it can be hard to know how words that you hear are written. Therefore, writing exercises help you to learn the relationship between sounds and letters in the language that you are learning.
If a speech sound does not occur in your language, you have to learn it. But even a sound that you have in your native language can be hard to pronounce. For example, some sounds may be easy to pronounce at the beginning of a word, but is not allowed at the end of a word in your native language: Chinese for instance allows only nasal consonants as in kin and king at the end of a syllable. So even though a Chinese has no problem with /p/ in pill, the same sound may be hard to pronounce at the end of the word lip. German and Dutch speakers will say /p/ and /s/ at the end of a syllable or word instead of /b/ and /z/, even though they allow both /b/ and /z/ at the beginning of a syllable/word. So even known sounds may be difficult in positions where they do not usually occur in your native language.
This is quite natural, actually. It is a result of the fact that your perception becomes more and more tuned to your native language in the first few months of life. This automatically also means that it becomes more and more difficult to distinguish categories in another language. If a new sound (category) is similar to a sound (category) in your native language, you will perceive it as the same, even if native speakers of the language you are learning can clearly and easily hear and pronounce the difference. This is called the “native language filter”. There are large individual differences, of course: Some people are better able to put the filter out of play than others.
Some languages allow very complex consonant clusters, like Polish. The word bezwzględny mean “ruthless” (quite appropriately for most leaners of Polish). Japanese, on the other hand, only allows syllables consisting of a single consonant followed by a vowel. Pronouncing unusual clusters can be hard, and we solve the problem by using different repair strategies: Language learners may simplify the clusters, substitute some sounds, add an e- at the beginning or insert a vowel into the cluster to break it up into two syllables; finally, we can reorder the consonants in the cluster. Since there is no knowing what learners may do (i.e., scientists cannot yet predict this on the basis of your native language), you get exercises for all repair strategies – until our logged user data are reliable enough to select only those exercises which deal with the repair strategy that other learners with the same native language use.
Prosody refers to properties of the language which are at a level higher than individual speech sounds. Word stress, lexical accents and intonation are prosodic properties. Since these are different across languages, CALST offers exercises for those. You can take the exercises several times, with different exercise material each time you take the exercises (except for the intonation exercises). This allows you to practise with new words as much as you like.
[Feature only available in Norwegian]
To answer your logical follow-up question “Why are there no exercises for prosody in English”: We are dependent on external funding for the implementation of exercises in English (and in other languages), which is hard to obtain. We hope to add these exercises in future, though.
In contrast to many other languages, Norwegian does not have a standard pronunciation. In their everyday use of the language, people speak their own dialect. Since these can be very different, you will have to learn to understand all of them when you learn Norwegian. So it is very useful to take the listening exercises for all the dialects in CALST. But you can choose one dialect which you want to learn to speak. This may be the dialect of the region where you are living, or often foreigners learn the Oslo dialect which is spoken by the largest group of Norwegians.

We are grateful for any questions and comments which help us to improve CALST. Please send an email to calst@hf.ntnu.no.

logo CALST

Logo. CALST

Logo. testCALST. Grafikk


How do the exercises work?

How do the exercises work?

Project members

User feedback

This is an awesome project! Thanks for making it! =)

CALST diverse

The project makes use of several resources developed by others:

Sound inventories

Sound inventory data are based on the UCLA Phonetic Segment Inventory Database (UPSID). These data are used to select sound contrast exercises dependent on the learner’s native language. This is done in the tool L1-L2map which lies behind CALST. We gratefully acknowledge the support of Ian Maddieson.

Pictures 

Pictures to illustrate the semantic content are an important part of the CALST program. The criteria we used as guidelines for the creation and selection of pictures were: drawings rather than photographs, simple style, few details, clear contours, intuitive, unambiguous, neutral and universal, but reflecting a "Norwegian reality" where relevant.

About one third of the pictures were taken from the open-source database UVic created by the University of Victoria, British Columbia.

The remaining pictures were either drawn on the computer or drawn by hand and subsequently scanned and edited by Eli Skarpnes and Egil Albertsen.

2016

Koreman, J., Bosoni, J.G., Abrahamsen, J. & Husby, O. (2016). L2 exercises for East-Norwegian word accents and intonation, Nordic Prosody XII, Trondheim (abstract).

Koreman, J., Albertsen, E., Martínez-Paricio, V., Husby, O. & Abrahamsen, J. (2016). Learning about word stress in L2 acquisition, New Sounds 2016, Aarhus (abstract).

Husby, O., Koreman, J., Albertsen, E. & Bosoni, J.G. (2016). Uttaleundervisning - progresjon, Den 7. nasjonale forskningskonferansen om norsk som andrespråk, Trondheim (abstract).

2015

Koreman, J., Martínez-Paricio, V., Abrahamsen, J. & Husby, O. (2015). A systematic approach to the pronunciation training of phonotactics, Proc. 18th Int. Congress of Phonetic Sciences (ICPhS2015), Glasgow.

Koreman, J., Husby, O., Hedayatfar, K. & Bech, Ø (2015). Learning from L2 learners to improve CAPT, ICPhS satellite workshop Phonetic Learner Corpora (abstract).

Martinez-Paricio, V. Koreman, J. Husby, O., Abrahamsen, J. & Bech, Ø. (2015). Consonant clusters in online L2 teaching: a multilingual approach, Proc. of the Pronunciation in Second Language Learning and Teaching (PSLLT) conference, volume 6, 115-125.

Martinez-Paricio, V. Koreman, J. Husby, O. (2015). L1-L2map: una base de datos fónica para la enseñanza de la pronunciación de segundas lenguasPerspectivas actuales en el análisis fónico del habla. Tradición y avances en la fonética experimental, 241-248. Publicacions de la Universitat de Valencia.

Husby, O., Koreman, J., Martínez-Paricio, V., Abrahamsen, J., Albertsen, E., Hedayatfar, K. & Bech, Ø (2015). Selective teaching of L2 pronunciation, in: Citical CALL. Proc. of the 2015 EUROCALL Conference, 243-248, Padova (abstract).

Koreman, J., Husby, O., Martínez-Paricio, V. & Abrahamsen, J. (2015). L1-L2map: una base de datos fónica para la enseñanza de la prononciación de segundas lenguas, in: A.C. Nebot (ed.), Perspectivas actuales en el análisis fónico del habla: tradición y avances en la fonética experimental, Departamento de Filología Española.

2014

Koreman, J., Husby, O., Martínez-Paricio, V. & Abrahamsen, J. (2014). Individuali­sering av læringsstien i norsk som andrespråk: lytte- og uttaletrening, Den 6. Nasjona­le Forskerkonferansen om Norsk som Andrespråk: Hvor Går Andrespråksforskningen?,Stavanger, Norway (abstract).

Martínez-Paricio, V., Koreman, J., Husby, O. (2014). CALST: Una herramienta digital para la enseñanza de la pronunciación de segundas lenguas, XV Congreso Internacio­nal de la Sociedad Española de Didáctica de la Lengua y la Literatura (SEDLL),Valencia, Spain (abstract).

Martínez-Paricio, V., Koreman, J., Husby, O., Abrahamsen, J.E. & Bech, Ø. (2014). Expanding CALST: multilingual analysis of L1-L2 phonotactics for language teachingPronunciation in Second Language Learning and Teaching (PSLLT 2014), Santa barbara, USA.

Web Content Display

 

 

FLERE PUBLIKASJONER CALST

 

2013

Koreman, J., Husby, O., Albertsen, E., Wik, P., Øvregaard, Å., Nefzaoui, S., Skarpnes, E. & Bech, Ø. (2013). Dealing with language diversity in teaching foreigners Norwegian pronunciationTromsø International Conference on Language Diversity, Tromsø (abstract).

Koreman, J., Husby, O., Albertsen, E., Wik, P., Øvregaard, Å., Nefzaoui, S., Skarpnes, E. & Bech, Ø. (2013). L1 variation in foreign language teaching: challenges and solutionsFifth Annual Conference on Pronunciation in Second Language Learning and Teaching, Ames, Iowa. (abstract).

Koreman, J., Wik, P., Husby, O. & Albertsen, E. (2013). Universal contrastive analysis as a learning principle in CAPTProc. of the workshop on Speech and Language Technology in Education (SLaTE 2013), ISSN 2311-4975, pp. 172-177.

Koreman, J. & Marinova, R. (2013). Språkteigen 17.02.2013, NRK P2 [Radio].

2012

Haugan, I. (2012). Treningsstudio for språkopplæring. Gemini 4, p. 6 (article in Nynorsk).

Koreman, J., Husby, O. & Wik, P. (2012). Comparing sound inventories for CAPT, in: O. Engwall (ed.), Proc. International Symposium on the Automatic Detection of Errors in Pronunciation Training (IS-ADEPT), 115-116. Stockholm: KTH, Computer Science and Communication (abstract).

2011

Husby, O., Øvregaard, Å., Wik, P., Bech, Ø., Albertsen, E., Nefzaoui, S., Skarpnes, E. & Koreman, J. (2011). Dealing with L1 background and L2 dialects in Norwegian CAPT, Proc. of the workshop on Speech and Language Technology in Education (SLaTE2011), Venice (Italy).

Koreman, J., Bech, Ø., Husby, O. & Wik, P. (2011). L1-L2map: a tool for multi-lingual contrastive analysis, Proc. 17th Int. Congress of Phonetic Sciences (ICPhS2011), Hong Kong.

Wik, Pr. Husby, O., Øvregaard, Å., Bech, Ø., Albertsen, E., Nefzaoui, S., Skarpnes, E. & Koreman, J. (2011). Contrastive analysis through L1-L2map. TRITA-TMH 2011, vol 51.

Øvregaard, Å. (2011). Språkteigen 19.06.2011, NRK P2 [Radio].

Øvregaard, Å. (2011). Språkteigen 27.11.2011, NRK P2 [Radio].

2010

Koreman, J. & Øvregaard, Å. (2010). Computer-Assisted Listening and Speaking Tutor, Høstseminar Norgesuniversitet 2010, Tromsø.

Koreman, J., Øvregaard, Å., Albertsen, E., Nefzaoui, S. & Skarpnes, E. (2010). Computer-Assisted Listening and Speaking Tutor, Individuell lytte- og uttaletrening med virtuell lærer. Nasjonal konferanse om bruk av IKT i utdanning og læring (NKUL 2010), Trondheim.

Øvregaard, Å. (2010). CALST (Computer-Assisted Listening and Speaking Tutor). NOAkonferansen, Voss.

Øvregaard, Å. (2010). CALST (Computer-Assisted Listening and Speaking Tutor). Hitech Lotech 2010.


2009

Koreman, J., Husby, O., Nefzaoui, S., Skarpnes, E., Øvregaard, Å. (2009). Computer-assisted Norwegian Teaching for Foreigners, Mutual Information Talks in ISK (MITISK), Institutt for språk og kommunikasjonsstudier NTNU, Trondheim.

2009: The CALST project started on 1 August 2009. The project was based on the Virtual Language Teacher (VILLE) an engine developed for Swedish by Preben Wik during his Ph.D. (see thesis) at the Division of Speech, Music and Hearing at KTH, Stockholm. Preben Wik was involved in the start of the project as part of the «think tank» and technical developer.

2013: The first version of CALST was launched by NTNU, Trondheim, on January 1. It was a downloadable version of the platform and contained vocabulary exercises for Norwegian and exercises for learning difficult Norwegian sounds based on the learner’s native language.

2015: The first web-based version appeared on January 1.

In April, a collaborative project to develop English contents was funded by the University of Agder (Allison Wetterlin, with Jacques Koreman).

A proposal for a Ph.D. project in NTNU’s Enabling Technologies initiative was granted. The doctoral candidate will apply feature-based automatic speech recognition (ASR) to automatic pronunciation evaluation. The proposal was written by Torbjørn Svendsen (IET) and Jacques Koreman (ISL). The project results can be used for automatic pronunciation evaluation in CALST.

On 1 July 2015, the exercises were extended with sound cluster exercises and a user logging was made implemented.

2017: In June 2017, the spin-off company Sounds Good was founded.

In October 2017, a new version of CALST, with a more user-friendly user interface, was launched.

The total budget for the CALST project is well around 15 million Norwegian krones.

The first CALST project was supported financially by Norway Opening Universities (Norgesuniversitetet) for the period from August 2009-August 2012.

NTNU, Faculty of Humanities as well as the Department of Language and Communication Studies have also contributed to the project, among other things by financing L1-L2map.

The webification of CALST and large parts of the following development of the platform were funded by VOX.

NTNU has contributed with financial support to the development of exercises for lexical tone and intonation.

Several ministries, amongst others Kommunal- og regionaldepartementet (KRD), Barne- og likestillingsdepartmentet (BLD) and Arbeidsdepartementet (AD), have funded further research in the CALST project through VOX, now Kompetanse Norge.

Spanish, Italian and Greek versions of CALST are being developed in the easyRights project (Horizon 2020 SC6-MIGRATION project number 870980, January 2020 – June 2023, see easyrights.eu).

Final evaluation by the steering committee: Wim van Dommelen, Helmer Strik and Björn Granström.

Introduction

During the years 2009-2012 a group of researchers and teachers of Norwegian as a second language at NTNU has developed a Computer-Assisted Listening and Speaking Tutor (CALST). The work was done in collaboration with the Centre for Speech Technology (CTT) at KTH (Stockholm), the Department of Linguistic and Scandinavian Studies (University of Oslo) and the Enhet for voksenopplæring (EVO, Trondheim).

The developers have succeeded in achieving the goals defined at the outset, viz. to build a system for teaching Norwegian. Therefore, to evaluate the CALST system, one can actually use the system itself.

The CALST system

The CALST system makes use of virtual animated language teachers that provide an audio-visual experience with both recorded natural utterances and visual animated articulation. The system can help beginning learners of Norwegian:

  1. to acquire a basic vocabulary with the help of a virtual teacher,
  2. to become familiar with the sounds of the Norwegian language, in particular sounds that are different or do not occur at all in the learner's native language,
  3. to practice the pronunciation of the sounds guided by the virtual teacher.

On the first screen the learner can specify his/her L1 and the desired dialect to learn. Then the corresponding L1-L2 map is loaded (see point 2 below).

Ad (1)

The present system contains a basic vocabulary of 1000 frequent words and expressions. The vocabulary is divided into various thematic categories. In the exercises words are pronounced by the virtual teacher and presented visually as small pictures. The student has to choose the appropriate picture and will get feedback. It is of great advantage that CALST offers a choice of four different dialects, each of those dialects being represented by a male and a female speaker as role models.

Ad (2)

The second component offers listening exercises for Norwegian sounds that may be difficult to distinguish for the particular learners (like the vowels in ‘by' vs. ‘bu'). What sounds the learner will experience as difficult depends on his or her native tongue. To be able to offer individual students exercises tailored to their language-specific needs, the tool L1-L2map was developed. This tool gives access to a database with information about speech sounds in 500 languages, making it a valuable asset for teachers and students.

Ad (3)

Using the system's third component, the learner can perform pronunciation exercises by imitating words spoken by the virtual teacher for the selected target dialect. This module offers the same four dialects as the vocabulary component. Apart from these ‘listen-and-repeat' exercises, there are ‘listen-and-write' exercises. Here, the student's task is to write down the word produced by the virtual teacher. The system gives the user feedback about the correctness of the written form.

To conclude, the goals mentioned in the proposal were achieved. In fact, even more than mentioned in the proposal has been achieved. Many learners have already used the system, and teachers also evaluated the system. Their useful feedback has been used to improve the system. In general, they were very positive about the system. The different components of the system work well, and the system is intuitive (not much instruction is needed to use it).

Future possibilities

Although the goals were met, and learners and teachers were positive, there are possibilities for further developments. Some of them are mentioned here.

Extra functionality could be added. For instance, something that was often mentioned by the users was automatic detection of, and feedback on pronunciation errors. This would make the system even more valuable for language learning. Furthermore, it might also be possible to integrate (parts of) the CALST system and the Norwegian on the Web (NoW) system.

The system might also be adapted for other target groups. The current CALST system already turned out to be useful for low- and non-literate learners. However, obviously, the system could be improved for this target group. This is interesting, since the number of low and non-literate learners is still quite substantial, they also generally lack sufficient opportunities for practising, esp. regarding pronunciation.

People with communicative disabilities (i.e. people with a language and/or speech handicap) also often need language training, such as pronunciation training. Again, also for this target group the system could be tuned, to make it better suited for them. Still, the current CALST system offers a good starting point for developing systems for these new groups.

Another possibility is to develop a CALL system for other target languages, other L2's. In that case the L1-L2map tool can still be used. Porting the system to another target language probably requires a substantially amount of work. However, such work can be carried out in cooperation with other partners, e.g. partners who already have expertise and resources for these languages.

The CALST system constitutes a nice starting point for a large amount of interesting, often interdisciplinary research. For developing such a system and for further developments, of which some possibilities were already described above, various kinds of expertise are needed, e.g. on linguistics, phonetics, (second, foreign) language acquisition, language learning, didactics, ICT, human-computer communication, ergonomics, etc. In some research projects, one could focus on one aspect, in another project on another aspect, making different kinds of grants, financing possible.

Finally, it is obvious that such language learning systems also have social and economic benefits, and that they are interesting form a commercial point of view. Every year, large amounts of people have to learn Norwegian, and other languages. In general, there are not enough opportunities to practise, esp. to practise oral communication. Such systems offer more possibilities for practise, since they are available 24/7, and learners can use to as much as they want. In many cases, using these systems can reduce the costs for learning languages.

All this opens up opportunities for cooperation with companies. Such cooperation with companies, valorisation, societal impact, etc., are becoming more and more important in research, and also for acquiring funding.

 

 

Thanks for providing this great resource! I will be excited to use the new version.

Jeg skal jobbe som lærer for syriske flyktninger fra høsten, og lurer på om det er mulig å få tilgang til CALST? Konseptet ser svært lovende ut, håper det passer for barn også.

Et supert opplegg!

Først, tusen takk for programmet ditt. Jeg lærte mye norsk når jeg fortsatt bodde i England ved å lese artikler og se på NRK (med undertekst), men det var CALST som hjalp meg utrolig mye med lytteforståelse. Jeg flyttet hit i september og besto Bergenstesten i oktober, og jeg kunne ikke ha gjort det uten dere.

Jeg jobber nå som engelsklærer og lurte på om dere visste om det var noe tilsvarende for engelsk. Det var ekstremt nyttig for meg i norsk, og jeg vet at elevene mine ville få den samme nytten av det.

This is an awesome project! Thanks for making it! =)

Your feedback encourages us to continue to improve CALST! Suggestions for improvements to calst@hf.ntnu.no are also welcome.