Features and Machine Learning Classification of Connected Speech Samples from Patients with Autopsy Proven Alzheimer's Disease with and without Additional Vascular Pathology

Abstract

Mixed vascular and Alzheimer-type dementia and pure Alzheimer's disease are both associated with changes in spoken language. These changes have, however, seldom been subjected to systematic comparison. In the present study, we analyzed language samples obtained during the course of a longitudinal clinical study from patients in whom one or other pathology was verified at post mortem. The aims of the study were twofold: first, to confirm the presence of differences in language produced by members of the two groups using quantitative methods of evaluation; and secondly to ascertain the most informative sources of variation between the groups. We adopted a computational approach to evaluate digitized transcripts of connected speech along a range of language-related dimensions. We then used machine learning text classification to assign the samples to one of the two pathological groups on the basis of these features. The classifiers' accuracies were tested using simple lexical features, syntactic features, and more complex statistical and information theory characteristics. Maximum accuracy was achieved when word occurrences and frequencies alone were used. Features based on syntactic and lexical complexity yielded lower discrimination scores, but all combinations of features showed significantly better performance than a baseline condition in which every transcript was assigned randomly to one of the two classes. The classification results illustrate the word content specific differences in the spoken language of the two groups. In addition, those with mixed pathology were found to exhibit a marked reduction in lexical variation and complexity compared to their pure AD counterparts.

Linguistic Features Identify Alzheimer’s Disease in Narrative Speech

“…Other researchers have suggested that 150 words is an acceptable minimum length for language analysis [36,78]. Our classification results indicate that there is still valuable information to be found in the short samples, confirming previous studies using picture descriptions to assess language in dementia [40,42,74]; however, we expect that the accuracies of each feature value would increase as the length of the sample increased. We also emphasize that our findings here do not necessarily generalize to other spoken language tasks.…”

“…Rentoumi et al [40] considered a slightly different problem: they used computational techniques to differentiate between picture descriptions from AD participants with and without additional vascular pathology (n = 18 for each group). They achieved an accuracy of up to 75% when they included frequency unigrams and excluded binary unigrams, syntactic complexity features, measures of vocabulary richness, and information theoretic features.…”

“…We call these features "key words", and they are integer-valued frequency counts. Previous work (e.g., [40,42]) has shown the utility of simple binary and frequency unigrams (word tokens). Rather than considering the space of all possible unigrams, we have considered only a smaller set which we have deemed to be relevant to the expected information content, to avoid problems of data sparsity, and to help improve the interpretability of the selected features.…”

Connected Speech in Neurodegenerative Language Disorders: A Review

“…Annotation and analysis of the text have been made using automated tools (i.e., part of speech tagger, parser), and machine learning methodology (support vector machine classifier, Bayesian networks, etc.). Machine Learning algorithms have been used to create diagnostic models using linguistic features resulting from speech samples (Garrard et al, 2014; Peintner et al, 2008; Jarrold et al, 2010, 2014; Guinn and Habash, 2012; Fraser et al, 2013, 2014a,b, 2015a,b; Rentoumi et al, 2014). In a few cases, the authors have also employed software for text analysis, namely the Linguistic Inquiry and Word Count, a tool that computes the frequency of words from predefined lists based on different categories such as psychological processes (i.e., emotional or cognitive), linguistic dimensions (i.e., articles, negations), or relativity (in time and space; Peintner et al, 2008; Jarrold et al, 2010, 2014).…”

Speech Recognition in Alzheimer’s Disease and in its Assessment

“…To ease this burden, modern approaches are being proposed to automate relevant aspects of assessment. Since AD can diminish vocabulary, syntactic complexity, and speech fluency, even in the earliest stages, various systems have been proposed to automatically detect signs of cognitive impairment from speech [2,3,4,5,6,7,8].…”