Skip to main content

Searching for linguistic signs of cognitive deterioration

Submitted by Kristina Lundholm Fors on 2019-11-01

In our research group, we are exploring ways of analysing language to find early signs of possible cognitive impairment, which may develop to dementia. Dementia is a condition that affects people all around the world, and the risk of developing it increases with age. There are several different types of dementia, but the most common one is Alzheimer's disease. This disease causes deterioration of cognitive functions, and one of the first symptoms tends to be problems with short-term memory. Language is also affected, and persons developing dementia experience difficulties with for example word finding, and may need more time to understand and process words and sentences. (Everyone, especially older persons, will sometimes experience short-term memory problems and word-finding difficulties, but in persons with Alzheimer's disease these problems are more severe and get worse over time.) There is currently no cure available for dementia, but actions can be taken to alleviate symptoms and introduce compensatory strategies for the person affected and their family. Therefore, it is important to identify signs of cognitive impairment at an early stage.

We have focused on trying to find signs of cognitive impairment in reading and spoken language, and we have compared persons with mild cognitive impairment and healthy controls. So far, the most promising results have come from analysing how the study participants move their eyes while reading. This was done using an eyetracker, but most likely this type of analysis while be possible to do with mobile phones in the future, making it more portable and available.

Foto på människa 
Collecting eyetracking data. The eyetracker is the device seen in front of the computer screen, and the headrest is used to help keep the head still, while performing the task.

Our study showed that persons with mild cognitive impairment tend to skip over words when reading and return to them later, while the controls have a greater tendency to read through the text from start-to-finish, and through using eye-tracking data we were able to distinguish between control participants and participants with MCI with at best 86% accuracy.

The manifestations of mild cognitive impairment in spoken language are subtle, but we have been able to show that linguistic analysis is helpful in identifying cognitive deterioration. For example, we have investigated our study participants' performance on the semantic verbal fluency test, which entails producing as many words from a specified category as possible during 60 seconds. (You can try that yourself: for example, name as many animals as possible during 60 seconds. It's harder than it sounds!) Typically, when doing this, persons produce words at a higher rate in the beginning of the minute, and start with common words. Later, less common words are retrieved, and as this requires more effort, the production rate is decreased. We analysed the test in 10-second intervals, and we could see that participants with MCI ran out of common words faster than the control group, and started producing less common words earlier (and thus their production rate slowed down).

To conclude, we are hopeful that linguistic analysis can become a valuable tool when screening for cognitive impairment, and the resources in the Språkbanken infrastructure are an important part of this, since they enable automatic analysis of different aspect of language.