Research

Research Interests

Language and speech science toward understanding human communication processes

+ as foundations of machinery spoken language processing and human language education
+ by modeling the mechanisms with data science techniques and linguistic analysis
+ sometimes with partners outside of our company (university laboratories etc.) 

Research Topics

 1. on exploring mechanisms

correct pronunciation: 

1) pitch accent fall prediction for unregistered Japanese words (family/first name) (including a kind of seq-to-seq approach) [Interspeech2003]
2) word reading disambiguation

coordinated prosodic expression:

1) analysis of prosodic event relating to naturalness (emphasis [SNLP 2009]) (modality [iSAI-NLP 2018] [ICCA2020])
2) prediction of prosodic event from linguistic information;
    + predicting phrase boundary rise in advertisement and conversation under the situation of synthesizing such a written text as flyer and novel [ Interspeech 2013]
    + predicting advertisement phrases to be emphasized [ Paclic 2014],
    + speech synthesis with the prosodic event labels [ Interspeech 2011, ICASSP 2013]

clear/easy-to-understand speech expression:

1) for elderly adults
    + analysis of desirable pause length in utterance when speaking with elderly adults [ASJ 2015 Autumn]
    + collection and analysis of speech data spoken by the exemplary speakers that elderly adults actually feel easy to understand [ASJ 2020 Spring] [Oriental COCOSDA 2020]


2. on preparing data for observation, analysis and machine learning use: 

data augmentation? / data collection by generation? :

1) for statistical language model [ Coling 2002]
2) for word reading disambiguation [ Paclic 2018][JSAI 2022]

corpus development:

for analyzing communicative speech compared to reading speech [Oriental COCOSDA 2010]

3. on questing for core {natural/spoken} language processing, dialogue processing and applying those technologies: