Automatic Proposition Extraction from Dependency Trees: Helping Early Prediction of Alzheimer's Disease from Narratives

Abstract

Abstract-Idea Density (ID) was originally proposed as a way of measuring the memory load of narratives, by representing the underlying content of the text as a series of semantic units, called propositions or ideas. From a clinical perspective, this notion has been shown to correlate with several cognitive aspects, such as memory, readability, aging, and dementia onset and progress. Traditionally, propositions are extracted manually from texts. There is a tool that can automate ID extraction [1], but it uses shallow information as input, and doesn't produce the propositions themselves as output. We propose a novel approach to obtaining the ID automatically from a text. Our method is an automation of Chand et al.'s ID manual [2], and consists of a rule-based system acting upon dependency trees. Initially, for each sentence in a text, a dependency parser is used to elicit the dependency relations between words. Then, a set of rules is recursively applied in order to process these relations to yield the corresponding propositions. We analyze preliminary results of our system using a well-formed journalistic text, and speech transcriptions of dementia patients.

Enriching Complex Networks with Word Embeddings for Detecting Mild Cognitive Impairment from Speech Transcripts

“…For the Portuguese language, machine learning algorithms were used to identify subjects with AD and MCI. Aluísio et al (2016) used a variety of linguistic metrics, such as syntactic complexity, idea density (da Cunha et al, 2015), and text cohesion through latent semantics. NLP tools with high precision are needed to compute these metrics, which is a problem for Portuguese since no robust dependency or constituency parsers exist.…”

Effects of regiolects on the perception of developmental foreign accent syndrome

“…BALE -Bateria de Avaliação da Linguagem no Envelhecimento (Hübner et al, in preparation) has been designed to assess semantic and discourse abilities in distinct educational levels in healthy aging and in neurodegenerative diseases. Automated discourse analysis tools have been applied to narrative transcripts in Brazilian Portuguese (Aluísio et al, 2016;Cunha et al, 2015). However, the absence of sentence boundary segmentation in transcripts prevents the direct application of methods that rely on these marks for the correct use of tools, such as taggers and parsers.…”

“…In this presentation, results from two discursive subtasks of BALE will be presented, one consisting of the retelling of an orally presented narrative, and the other a narrative production based on a sequence of scenes. Data analyses are performed by DeepBond, a method of automatic sentence segmentation (Treviso et al, 2017), together with the use of Coh-Metrix Dementia (CMD) (Aluísio et al, 2016;Cunha et al, 2015), which provides 73 metrics, including syntactic complexity. Moreover, data from questions measuring micro and macrostructure comprehension and text propositions are analyzed.…”

Psycholinguistic Models of Sentence Processing Improve Sentence Readability Ranking

“…These features comprise the INTEGRATIONCOST model. Finally, we use a modified version of the IDD3 library from Andre Cunha (Cunha et al, 2015) to extract idea density decomposed across three types of propositional idea: predications, modifications, and connections. 10 Here we use only averaged features, as the crucial measure is the idea density rather than the raw number of ideas being expressed.…”