Recently, dementia patients have been increasing in number worldwide, necessitating the development of techniques to detect dementia as early as possible. Considering that a typical symptom of dementia, especially Alzheimer’s disease, is language impairment, speech-based dementia detection approaches have drawn much attention. This paper presents a smartphone-based dementia screening application, VocabChecker, which measures language abilities from a speech narrative via automatic speech recognition (ASR). It measures four language abilities related to dementia: number of tokens (token), number of types (type), type token ratio (TTR), and potential vocabulary size (PVS). We also reported that the use of VocabChecker has distinguished dementia patients from elderly people.