Case Kontula: Multilingual map survey

In the Kontula Mall pilot project, we developed a multilingual map survey for foreign language respondents using the Maptionnaire tool. Maptionnaire is an e-participation tool widely used in urban planning. It has also been used by the City of Helsinki. We chose this tool precisely because it is a familiar and well-established tool in the city. However, we applied the questionnaire in a completely new way: the content was designed with the user in mind and translated into foreign languages we knew were spoken by the shopkeepers. These languages were Arabic, Bengali, English, Kurdish (Sorani), and Turkish.

The first thing to do when designing a multilingual survey is to identify the languages for which translations are needed. These languages need to be taken into account not only in the production of the survey content, but also in the languages of the survey tool. As a result of our experimentation, Bengali and Kurdish were added as new languages to the Maptionnaire tool. The addition of languages to the tool was agile.

The practical experiment of designing and implementing a multilingual survey taught me a lot. First, all languages in a multilingual questionnaire must have equal status. However, the survey tool was designed with the idea that there should always be one main language. We solved this problem by creating a front page that guides respondents to select a language before moving on.

Multilingual translations also present their own challenges. There were several language interpreters, some translated from Finnish and some from English into other languages. As the Maptionnaire translation tool was built around the main language of the survey, we decided to compile the translations into a separate Excel file, where translations between languages could be done. As the author of the questionnaire, I experienced many confusing moments when copying foreign language texts into the questionnaire. I hope I was careful not to make any copying mistakes, as I did not understand the languages myself. I felt like I was acting blind.

The Arabic and Kurdish languages, where the text runs from right to left, added to the challenge. Maptionnaire was good at making the alignments, but as someone used to Western spelling, I found it difficult to handle the texts.

The translation tools Google Translate and DeepL were in heavy use. They helped me, even with no knowledge of the language, to get an idea of the meaning of the texts. Some corrections to the questionnaire were made after the translations were completed and had to be translated using these tools. Fortunately, we could check the translations we had made with the interpreters when the questionnaire started to be used.

We also encountered the challenges of multilingualism when conducting the interviews. The Kurdish interpreter could not write down the interviewee’s answers to the questionnaire, so the Finnish researcher had to write the interpreted answers in Finnish on the Kurdish version of the questionnaire – and here the text went from right to left…

In the analysis phase, we were again faced with translation issues. Our questionnaire data consisted of responses in Turkish, Finnish, and English. The Turkish text was translated into Finnish by Google Translate and checked by an interpreter.

From a multilingual perspective, the development of survey tools could, in the future, also consider translating the response material. Of course, translations can be made from raw data, but ideally, they would be exported to the system in the same way as the user interface translations.

As for digitalization, I hope automatic translation tools will become so good that their translations can be trusted. I also look forward to integrating automatic translations into the survey tools on the maintenance side and in the respondent interface. I believe that one day, we can easily read content in any language translated into our own language.

This is an English version of Pilvi’s blog, which you can find here.