Daichi Mochihashi

The Institute of Statistical Mathematics Department of Statistical Inference and Mathematics Associate Professor

Message

I am looking forward to working with the team.

Research Plan Overview

言語にPrevious studies on language have primarily focused on the content the language communicates, primarily because language was originally developed as a means of survival. However, in civilizations where survival is guaranteed, people develop language not only to communicate objective information but also to share subjective affects, primarily in the form of literary art such as songs and poems. To take a scientific approach on such issues regarding affect, this project aims to reveal (A) what affects are evoked in the brain by language and (B) what language expressions and structures are responsible for the different affects through neuroscientific, linguistic, and natural language processing methods.

From an engineering perspective, this project’s findings may enable people to predict the affective impressions created by songs or novels and generate advertisement catchphrases that evoke certain desirable affects, based on how the brain functions.

Experimental results (ongoing)
・Interpreting regions of brain activity with Neuro synth
・Activity related to the language cortex can be observed

The background context and objectives of this research project

Previous studies on language, in the fields of linguistics and natural language processing/computational linguistics, have primarily focused on the content the language communicates. This is because language was originally developed as a means of communicating essential information for survival and communal living (Okanoya, 2003). However, in civilizations where survival is mostly guaranteed, people develop language methods not only to communicate objective information for survival but also to share subjective affects. These include forms of expression such as paintings (not as a means of expressing information for survival), music, and dance, while typical examples of literary art would be songs and poems. The purpose of such literary art forms is not to communicate an objective description of the external world but rather to express and convey one’s subjective affects through the art.

That being said, it is difficult to interpret such affects from a literary text. This is because descriptions of affect are inextricably linked to one’s personal experiences and the tangible elements of language cannot be conveyed through text alone. In the sentence “clammbon wa kapukapu waratta yo” (Kenji Miyazawa's Yamanashi), for instance, the emotional nuance of the term “clammbon” or the feeling that is evoked by the sound “kapukapu” cannot be discovered by simply analyzing the text. In the haiku “kangyōno minahitorinaru shihatsukana” (Yu Kaneshiro), there are no descriptions of the various affects the readers feel through personal experience as well.

Thus, in order to scientifically approach such issues involving affect, one must collect data from the organ that is responsible for evoking affect, that is, the brain. This project aims to reveal (A) what affects are evoked in the brain by language and (B) what language expressions and structures are responsible for the different affects through neuroscientific, linguistic, and natural language processing methods. By doing so, the project hopes to objectify the affective function of language based on data, something that conventional (computational) linguistics has not achieved. This project, therefore, can be considered to be a quantitative approach to literary art. From an engineering perspective, this project’s findings may enable people to predict the affective impressions created by songs or novels and generate advertisement catchphrases that evoke certain desirable affects, based on how the brain functions.

References