Who does what to whom? graph representations of action-predication in speech relate to psychopathological dimensions of psychosis

Abstract

Graphical representations of speech generate powerful computational measures related to psychosis. Previous studies have mostly relied on structural relations between words as the basis of graph formation, i.e., connecting each word to the next in a sequence of words. Here, we introduced a method of graph formation grounded in semantic relationships by identifying elements that act upon each other (action relation) and the contents of those actions (predication relation). Speech from picture descriptions and open-ended narrative tasks were collected from a cross-diagnostic group of healthy volunteers and people with psychotic or non-psychotic disorders. Recordings were transcribed and underwent automated language processing, including semantic role labeling to identify action and predication relations. Structural and semantic graph features were computed using static and dynamic (moving-window) techniques. Compared to structural graphs, semantic graphs were more strongly correlated with dimensional psychosis symptoms. Dynamic features also outperformed static features, and samples from picture descriptions yielded larger effect sizes than narrative responses for psychosis diagnoses and symptom dimensions. Overall, semantic graphs captured unique and clinically meaningful information about psychosis and related symptom dimensions. These features, particularly when derived from semi-structured tasks using dynamic measurement, are meaningful additions to the repertoire of computational linguistic methods in psychiatry.

Automatic evaluation-feedback system for automated social skills training

“…Another challenge is capturing an appropriate representation of the speech contents. To deal with this problem, such cutting-edge NLP methods as speech-graph representations are required 45 , 46 .…”

Evolution of linguistic markers of agency, centrality and content during metacognitive therapy for psychosis

“…Next, we generated semantic graph models to determine the activity/passivity, and centrality of the pronoun categories. 32 All features are enlisted in Table 1. Speech graph methodology is illustrated in Figure 1.…”

“…To normalize for the effect of verbosity, we used dynamic measurements with a moving window of thirty predicative blocks and forwardmoving steps of one predicate. 32 Action-predication graphs were generated by connecting predicates to all arguments (A0, A1, and A2), and active arguments (A0) to passive arguments (A1 and A2) in each predicative block in an undirected fashion to allow centrality computation (e.g., 'bring'−'they', 'bring'−'joy', 'bring'−'I life', 'they'−'joy' and 'they'−'I life'). Based on this graph model, we then computed betweenness centralities of nodes, i.e., how often a particular node mediates the shortest path between other nodes.…”

“…First, semantic role labelling is able to identify and quantify agency in terms of active or passive linguistic relations. 32 For example, while "Good things always happen to me" and "I bring joy to my life" both have positive connotations, the latter utterance is associated with more sense of agency as reflected in the active role acquired by the first-person pronoun. An increase in the use of active roles for the first-person pronoun during therapy sessions has been shown to be associated with good psychotherapy outcomes.…”