Translating Sound Adjectives by Collectively Teaching Abstract Representations

Cartwright, M., Pardo, B. Translating Sound Adjectives by Collectively Teaching Abstract Representations. In Proceedings of the Collective Intelligence Conference, 2014.


Correctly translating adjectives that describe sound (e.g. "heavy", "tinny", "dry", "soothing") can be a difficult task. Resources such as the Oxford English Dictionary (OED) [Oxford University Press 1992], typically list the "audio sense” for only a small subset of the words commonly used to describe sound. For example, "warm” is a very commonly used sound adjective and the OED does not mention the audio sense. Directly translating the predominant (i.e. first) sense of a sound adjective into another language often results in an incorrect translation. For example, when "a warm sound" is typed into Google Translate, it responds with "un sonido c ´alido.” While the word-for-word translation is correct, the appropriate translation to correctly express the meaning is "un sonido profundo." The word-for-word translation to English of "un sonido profundo" is "a deep sound," not "a warm sound." As a result, people relying on current translation technology may fail to communicate while believing they have. This, for example, would make it difficult for an English-speaking audiologist to correctly diagnose hearing problems for people whose primary language is not English.

Here, we describe a system that builds a translation map between sound adjectives of two languages: English and Spanish. This map is built from the collective intelligence of hundreds of participants who teach the system sound adjectives by indicating how well example sounds embody the adjectives. When two words are both strongly embodied by the same sound examples, they are considered synonyms. When the two words come from different languages, we consider one a translation of the other. The more frequently a pairing between two words occurs, the more certain the translation.