Some lucky students, attending the Korean language and culture course at the University for Foreigners of Siena, had the privilege of being able to participate in a project on Korean language testing.

This academic year some lucky students, attending the Korean language and culture course at the University for Foreigners of Siena, had the privilege of being able to participate in a project on Korean language testing, which is financed by the National Information society Agency (NIA) of the South Korean government.

The Test of Proficiency in Korean (TOPIK) is a Korean language test for non-native speakers, instituted in 1997 by the Ministry of Education. Over the years the test has undergone several substantial changes, starting with the subdivision into 2 groups instead of 3 and the addition of the writing part in TOPIK II. Recently, the National Institute for International Education has proposed to introduce the speaking part as well, given that the number of people eager for a Korean language certification is constantly growing.

The institutions have determined that by 2022 the evaluation will be done by computer with an artificial intelligence program created to facilitate the language assessment. The program will recognize the pronunciation and speech ability of the speaker, giving questions for pronunciation, listening and repetition, short answers, and free answers.

With the collaboration of several Korean universities and professors, last year they began collecting a database of sounds, words, phrases, and speeches to develop this new technology. Initially, the first people to record their voices were Korean native speakers, but when foreign speakers began to participate, they all realized a fundamental problem: the accent and pronunciation differ depending on the speaker’s mother tongue. 

The language learning abilities of human beings vary due to many different factors, but regarding phonetic and pronunciation ability, scholars have proven that there is a critical learning threshold of around 12-14 years. The older you grow, the harder it is to be able to pronounce sounds other than those of your own language. 

When English speakers and Japanese speakers tried to pronounce the word “McDonald”, the artificial intelligence did not recognize their accent and pronunciation and considered it a mistake. Having encountered this problem several times, the researchers decided to start the “Korean speech database construction project for learning through AI”, in which several universities and schools from all over the world were invited to participate.

According to this project, each volunteer needs to record words, phrases, and answers naturally on the AI platform using a profile created with the credentials provided by the company. This way, the artificial intelligence memorizes all the different accents and pronunciations in order to recognize them in the future. When the program is used during speech evaluation in TOPIK II, words and phrases with a slight accent will not be classified as errors.

The University for Foreigners of Siena was the first Italian university to be involved and many students have volunteered, eager to participate believing that by using this technology, we could have more impartial, precise, and certainly less time-consuming evaluations.   

Ilaria Righi is an Italian student at the University for Foreigners of Siena, majoring in Cultural Mediation with a specialization in English translation and Korean studies. She has recently completed the spring semester as an international exchange student at Ewha Womans University in Seoul. – Ed