Search results

Computerized assessment of the acoustics of progressive aphasia

From the introduction: "There are two variants of fronto-temporal dementia: a behavioral variant (behavioral FTD, bvFTD, Neary et al. (1998)), which causes changes in behavior and personality but leaves syntax, phonology and semantics relatively intact, and a variant that causes impairments in the language processing system (Primary Progessive Aphasia, PPA (Gorno-Tempini et al., 2004). PPA can be subdivided into subtypes fluent (fluent but empty speech, comprehension of word meaning is affected / `semantic dementia') and non-fluent (agrammatism, hesitant or labored speech, word finding problems). Some identify logopenic aphasia as a FTD-variant: fluent aphasia with anomia but intact object recognition and underlying word meaning."

Exploring Bias in Data and Models for Misinformation Detection from Text

With the proliferation of misinformation on the web, automatic misinformation detection methods are becoming an increasingly important subject of study. Large language models have produced the best results among content-based methods, which rely on the text of the article rather than the metadata or network features. However, finetuning such a model requires significant training data, which has led to the automatic creation of large-scale misinformation detection datasets. In these datasets, articles are not labelled directly. Rather, each news site is labelled for reliability by an established fact-checking organisation and every article is subsequently assigned the corresponding label based on the reliability score of the news source in question. A recent paper has explored the biases present in one such dataset, NELA-GT-2018, and shown that the models are at least partly learning the stylistic and other features of different news sources rather than the features of unreliable news. We confirm a part of their findings. Apart from studying the characteristics and potential biases of the datasets, we also find it important to examine in what way the model architecture influences the results. We therefore explore which text features or combinations of features are learned by models based on contextual word embeddings as opposed to basic bag-of-words models. To elucidate this, we perform extensive error analysis aided by the SHAP post-hoc explanation technique on a debiased portion of the dataset. We validate the explanation technique on our inherently interpretable baseline model.

The essential elements for a nursing home according to stakeholders from healthcare and technology

Technology and architectural solutions are needed as a means of support in future nursing homes. This study investigated how various monodisciplinary groups of stakeholders from healthcare and technology envision the nursing home of the future and which elements are necessary for its creation. Moreover, differences in needs and interests between the various stakeholders were considered. This qualitative study gathered data via 10 simultaneous sticky note brainstorm sessions with 95 professional stakeholders, which resulted in 1459 quotes in five categories that were clustered into themes and processed into word clouds. The stakeholders prioritized the needs of the resident and placed the most importance on the fact that a nursing home is primarily a place to live in the final stages of one's life. A mix of factors related to the quality of care and the quality of the built environment and technology is needed. Given the fact that there are differences in what monodisciplinary groups of stakeholders see as an ideal nursing home, multidisciplinary approaches should be pursued in practice to incorporate as many new views and stakeholder needs as possible.

Eindrapportage DNA-Succesmeter

Sentiment polarity classification of corporate review data with a bidirectional Long-Short Term Memory (biLSTM) neural network architecture

A considerable amount of literature has been published on Corporate Reputation, Branding and Brand Image. These studies are extensive and focus particularly on questionnaires and statistical analysis. Although extensive research has been carried out, no single study was found which attempted to predict corporate reputation performance based on data collected from media sources. To perform this task, a biLSTM Neural Network extended with attention mechanism was utilized. The advantages of this architecture are that it obtains excellent performance for NLP tasks. The state-of-the-art designed model achieves highly competitive results, F1 scores around 72%, accuracy of 92% and loss around 20%.

Using brain potentials to functionally localise stroop-like effects in colour and picture naming : perceptual encoding versus word planning.

The colour-word Stroop task and the picture-word interference task (PWI) have been used extensively to study the functional processes underlying spoken word production. One of the consistent behavioural effects in both tasks is the Stroop-like effect: The reaction time (RT) is longer on incongruent trials than on congruent trials. The effect in the Stroop task is usually linked to word planning, whereas the effect in the PWI task is associated with either word planning or perceptual encoding. To adjudicate between the word planning and perceptual encoding accounts of the effect in PWI, we conducted an EEG experiment consisting of three tasks: a standard colour-word Stroop task (three colours), a standard PWI task (39 pictures), and a Stroop-like version of the PWI task (three pictures). Participants overtly named the colours and pictures while their EEG was recorded. A Stroop-like effect in RTs was observed in all three tasks. ERPs at centro-parietal sensors started to deflect negatively for incongruent relative to congruent stimuli around 350 ms after stimulus onset for the Stroop, Stroop-like PWI, and the Standard PWI tasks: an N400 effect. No early differences were found in the PWI tasks. The onset of the Stroop-like effect at about 350 ms in all three tasks links the effect to word planning rather than perceptual encoding, which has been estimated in the literature to be finished around 200-250 ms after stimulus onset. We conclude that the Stroop-like effect arises during word planning in both Stroop and PWI.