Four dimensions of naturalistic language production in aphasia after stroke.

There is a rich tradition of research on the neuroanatomical correlates of spoken language production in aphasia using constrained tasks (e.g., picture naming), which offer controlled insights into the distinct processes that govern speech and language (i.e., lexical-semantic access, morphosyntactic construction, phonological encoding, speech motor programming/execution). Yet these tasks do not necessarily reflect everyday language use. In contrast, naturalistic language production (also referred to as connected speech or discourse) more closely approximates typical processing demands, requiring the dynamic integration of all aspects of speech and language. The brain bases of naturalistic language production remain relatively unknown, however, in part because of the difficulty in deriving features that are salient, quantifiable, and interpretable relative to both speech-language processes and the extant literature. The present cross-sectional observational study seeks to address these challenges by leveraging a validated and comprehensive auditory-perceptual measurement system that yields four explanatory dimensions of performance-Paraphasia (misselection of words and sounds), Logopenia (paucity of words), Agrammatism (grammatical omissions), and Motor speech (impaired speech motor programming/execution). We used this system to characterize naturalistic language production in a large and representative sample of individuals with acute post-stroke aphasia (n = 118). Scores on each of the four dimensions were correlated with lesion metrics, and multivariate associations among the dimensions and brain regions were then explored. Our findings revealed distinct yet overlapping neuroanatomical correlates throughout the left-hemisphere language network. Paraphasia and Logopenia were associated primarily with posterior regions, spanning both dorsal and ventral streams, which are critical for lexical-semantic access and phonological encoding. In contrast, Agrammatism and Motor speech were associated primarily with anterior regions of the dorsal stream that are involved in morphosyntactic construction and speech motor planning/execution respectively. Collectively, we view these results as constituting a brain-behavior model of naturalistic language production in aphasia, aligning with both historical and contemporary accounts of the neurobiology of spoken language production.

Naturalistic Spoken Language Comprehension Is Supported by Alpha and Beta Oscillations.

Brain oscillations are prevalent in all species and are involved in numerous perceptual operations. α oscillations are thought to facilitate processing through the inhibition of task-irrelevant networks, while ß oscillations are linked to the putative reactivation of content representations. Can the proposed functional role of α and ß oscillations be generalized from low-level operations to higher-level cognitive processes? Here we address this question focusing on naturalistic spoken language comprehension. Twenty-two (18 female) Dutch native speakers listened to stories in Dutch and French while MEG was recorded. We used dependency parsing to identify three dependency states at each word: the number of (1) newly opened dependencies, (2) dependencies that remained open, and (3) resolved dependencies. We then constructed forward models to predict α and ß power from the dependency features. Results showed that dependency features predict α and ß power in language-related regions beyond low-level linguistic features. Left temporal, fundamental language regions are involved in language comprehension in α, while frontal and parietal, higher-order language regions, and motor regions are involved in ß. Critically, α- and ß-band dynamics seem to subserve language comprehension tapping into syntactic structure building and semantic composition by providing low-level mechanistic operations for inhibition and reactivation processes. Because of the temporal similarity of the α-ß responses, their potential functional dissociation remains to be elucidated. Overall, this study sheds light on the role of α and ß oscillations during naturalistic spoken language comprehension, providing evidence for the generalizability of these dynamics from perceptual to complex linguistic processes.SIGNIFICANCE STATEMENT It remains unclear whether the proposed functional role of α and ß oscillations in perceptual and motor function is generalizable to higher-level cognitive processes, such as spoken language comprehension. We found that syntactic features predict α and ß power in language-related regions beyond low-level linguistic features when listening to naturalistic speech in a known language. We offer experimental findings that integrate a neuroscientific framework on the role of brain oscillations as "building blocks" with spoken language comprehension. This supports the view of a domain-general role of oscillations across the hierarchy of cognitive functions, from low-level sensory operations to abstract linguistic processes.

Linguistic network in early deaf individuals: A neuroimaging meta-analysis.

This meta-analysis summarizes evidence from 44 neuroimaging experiments and characterizes the general linguistic network in early deaf individuals. Meta-analytic comparisons with hearing individuals found that a specific set of regions (in particular the left inferior frontal gyrus and posterior middle temporal gyrus) participates in supramodal language processing. In addition to previously described modality-specific differences, the present study showed that the left calcarine gyrus and the right caudate were additionally recruited in deaf compared with hearing individuals. In addition, this study showed that the bilateral posterior superior temporal gyrus is shaped by cross-modal plasticity, whereas the left frontotemporal areas are shaped by early language experience. Although an overall left-lateralized pattern for language processing was observed in the early deaf individuals, regional lateralization was altered in the inferior frontal gyrus and anterior temporal lobe. These findings indicate that the core language network functions in a modality-independent manner, and provide a foundation for determining the contributions of sensory and linguistic experiences in shaping the neural bases of language processing.

The Roles of Segments and Tone in Mandarin Lexical Processing: An ERP Study.

BACKGROUNDS: Segments and tone are important sub-syllabic units that play large roles in lexical processing in tonal languages. However, their roles in lexical processing remain unclear, and the event-related potential (ERP) technique will benefit the exploration of the cognitive mechanism in lexical processing. METHODS: The high temporal resolution of ERP enables the technique to interpret rapidly changing spoken language performances. The present ERP study examined the different roles of segments and tone in Mandarin Chinese lexical processing. An auditory priming experiment was designed that included five types of priming stimuli: consonant mismatch, vowel mismatch, tone mismatch, unrelated mismatch, and identity. Participants were asked to judge whether the target of the prime-target pair was a real Mandarin disyllabic word or not. RESULTS: Behavioral results including reaction time and response accuracy and ERP results were collected. Results were different from those of previous studies that showed the dominant role of consonants in lexical access in mainly non-tonal languages like English. Our results showed that consonants and vowels play comparable roles, whereas tone plays a less important role than do consonants and vowels in lexical processing in Mandarin. CONCLUSIONS: These results have implications for understanding the brain mechanisms in lexical processing of tonal languages.

The everyday speech environments of preschoolers with and without cochlear implants.

Children who receive cochlear implants develop spoken language on a protracted timescale. The home environment facilitates speech-language development, yet it is relatively unknown how the environment differs between children with cochlear implants and typical hearing. We matched eighteen preschoolers with implants (31-65 months) to two groups of children with typical hearing: by chronological age and hearing age. Each child completed a long-form, naturalistic audio recording of their home environment (appx. 16 hours/child; >730 hours of observation) to measure adult speech input, child vocal productivity, and caregiver-child interaction. Results showed that children with cochlear implants and typical hearing were exposed to and engaged in similar amounts of spoken language with caregivers. However, the home environment did not reflect developmental stages as closely for children with implants, or predict their speech outcomes as strongly. Home-based speech-language interventions should focus on the unique input-outcome relationships for this group of children with hearing loss.

Map Task Corpus of Heritage BCMS spoken by second-generation speakers in Switzerland.

In this paper, we present a corpus for heritage Bosnian/Croatian/Montenegrin/Serbian (BCMS) spoken in German-speaking Switzerland. The corpus consists of elicited conversations between 29 second-generation speakers originating from different regions of former Yugoslavia. In total, the corpus contains 30 turn-aligned transcripts with an average length of 6 min. It is enriched with extensive speakers' metadata, annotations, and pre-calculated corpus counts. The corpus can be accessed through an interactive corpus platform that allows for browsing, querying, and filtering, but also for creating and sharing custom annotations. Principal user groups we address with this corpus are researchers of heritage BCMS, as well as students and teachers of BCMS living in diaspora. In addition to introducing the corpus platform and the workflows we adopted to create it, we also present a case study on BCMS spoken by a pair of siblings who participated in the map task, and discuss advantages and challenges of using this corpus platform for linguistic research.

Pre-Trained Joint Model for Intent Classification and Slot Filling with Semantic Feature Fusion.

The comprehension of spoken language is a crucial aspect of dialogue systems, encompassing two fundamental tasks: intent classification and slot filling. Currently, the joint modeling approach for these two tasks has emerged as the dominant method in spoken language understanding modeling. However, the existing joint models have limitations in terms of their relevancy and utilization of contextual semantic features between the multiple tasks. To address these limitations, a joint model based on BERT and semantic fusion (JMBSF) is proposed. The model employs pre-trained BERT to extract semantic features and utilizes semantic fusion to associate and integrate this information. The results of experiments on two benchmark datasets, ATIS and Snips, in spoken language comprehension demonstrate that the proposed JMBSF model attains 98.80% and 99.71% intent classification accuracy, 98.25% and 97.24% slot-filling F1-score, and 93.40% and 93.57% sentence accuracy, respectively. These results reveal a significant improvement compared to other joint models. Furthermore, comprehensive ablation studies affirm the effectiveness of each component in the design of JMBSF.

Linguistic Abnormalities in Isolated Rapid Eye Movement Sleep Behavior Disorder.

BACKGROUND: Patients with synucleinopathies frequently display language abnormalities. However, whether patients with isolated rapid eye movement sleep behavior disorder (iRBD) have prodromal language impairment remains unknown. OBJECTIVE: We examined whether the linguistic abnormalities in iRBD can serve as potential biomarkers for conversion to synucleinopathy, including the possible effect of mild cognitive impairment (MCI), speaking task, and automation of analysis procedure. METHODS: We enrolled 139 Czech native participants, including 40 iRBD without MCI and 14 iRBD with MCI, compared with 40 PD without MCI, 15 PD with MCI, and 30 healthy control subjects. Spontaneous discourse and story-tale narrative were transcribed and linguistically annotated. A quantitative analysis was performed computing three linguistic features. Human annotations were compared with fully automated annotations. RESULTS: Compared with control subjects, patients with iRBD showed poorer content density, reflecting the reduction of content words and modifiers. Both PD and iRBD subgroups with MCI manifested less occurrence of unique words and a higher number of n-grams repetitions, indicating poorer lexical richness. The spontaneous discourse task demonstrated language impairment in iRBD without MCI with an area under the curve of 0.72, while the story-tale narrative task better reflected the presence of MCI, discriminating both PD and iRBD subgroups with MCI from control subjects with an area under the curve of up to 0.81. A strong correlation between manually and automatically computed results was achieved. CONCLUSIONS: Linguistic features might provide a reliable automated method for detecting cognitive decline caused by prodromal neurodegeneration in subjects with iRBD, providing critical outcomes for future therapeutic trials. © 2022 International Parkinson and Movement Disorder Society.

Convergent validity of functional communication tools and spoken language comprehension assessment in children with cerebral palsy.

BACKGROUND: The majority of children with cerebral palsy (CP) experience challenges in functional communication from a young age. A pivotal aspect of functional communication is language comprehension. A variety of classification systems and questionnaires are available to classify and describe functional communication skills in children with CP. A better understanding of the convergent validity of (subsections of) these tools, as well as their relationship with spoken language comprehension, will be valuable in both clinical practice and research. AIMS: To investigate the convergent validity of (subsections of) functional communication tools and the relationship with spoken language comprehension in children with CP. METHODS & PROCEDURES: Cross-sectional data on 138 children were subdivided into three developmental stages based on (Dutch) educational phases: ages 18 months-3;11y (n = 59), 4;0-5;11 years (n = 37) and 6;0-8;11 years (n = 42). The following functional communication tools were used to classify and describe functional communication: Communication Function Classification System (CFCS), subscales of the Caregivers Priorities and Child Health Index of Life with Disabilities-Dutch Version (CPCHILD-DV) and the Focus on Communication Under Six-34 (FOCUS-34) questionnaire. Spoken language comprehension was assessed with the Computer-Based instrument for Low motor Language Testing (C-BiLLT). Correlations between the functional communication tools, and with the C-BiLLT, were calculated using Pearson's and Spearman's correlation coefficients. It was hypothesized a priori that correlations of at least 0.60 suggest good convergent validity. OUTCOMES & RESULTS: At all developmental stages, a significant ordered decreasing tendency of communication outcomes was found across CFCS levels; lower CFCS levels were associated with lower scores on the CPCHILD-DV and FOCUS-34, and with a lower level of spoken language comprehension (C-BiLLT). Correlation coefficients of the functional communication tools exceeded 0.60 at all developmental stages. Correlations between C-BiLLT raw scores and the functional communication tools varied between 0.351 and 0.591 at developmental stage 18 months-3;11 years, between 0.781 and 0.897 at developmental stage 4;0-5;11 years, and between 0.635 and 0.659 at developmental stage 6;0-8;11 years. CONCLUSIONS & IMPLICATIONS: The functional communication tools assessed in this study showed convergent validity at all developmental stages. The CFCS, currently most widely used in paediatric rehabilitation, is adequate in the classification of functional communication. However, for more detailed clinical goal setting and evaluation of change in functional communication, the additional use of FOCUS-34 or CPCHILD-DV is recommended. WHAT THIS PAPER ADDS: What is already known on the subject A range of functional communication tools are available that help describe and classify functional communication in children with CP. These include the CFCS, subsections of CPCHILD-DV and FOCUS-34. The CFCS classifies functional communication in daily life with familiar and unfamiliar partners. Specific subsections of the CPCHILD-DV and FOCUS-34 include items that pertain to communicative participation. The innovative C-BiLLT provides a standardized method to assess spoken language comprehension in children with CP and significant motor impairments. What this paper adds to existing knowledge In the present study, convergent validity was confirmed between CFCS and specific subsections of the CPCHILD-DV and FOCUS-34. Correlations between these functional communication tools and the C-BiLLT were moderate to strong. What are the potential or actual clinical implications of this work? For clinical and research purposes (for instance, accurate prescription of augmentative and alternative communication-AAC), healthcare and educational professionals together with parents need to know how functional communication tools converge and how functional communication levels relate to the comprehension of spoken language. The CFCS provides a valid classification of functional communication abilities in children with CP. However, to measure change in functional communication and to evaluate treatment outcomes, use of additional functional communication tools such as the CPCHILD-DV and FOCUS-34 is recommended. When discrepancies are found between communicative abilities and spoken language comprehension, it is strongly recommended that valid tools are used in a more detailed examination of the child's spoken language comprehension skills and functional communication.

Improved Spoken Language Representation for Intent Understanding in a Task-Oriented Dialogue System.

Successful applications of deep learning technologies in the natural language processing domain have improved text-based intent classifications. However, in practical spoken dialogue applications, the users' articulation styles and background noises cause automatic speech recognition (ASR) errors, and these may lead language models to misclassify users' intents. To overcome the limited performance of the intent classification task in the spoken dialogue system, we propose a novel approach that jointly uses both recognized text obtained by the ASR model and a given labeled text. In the evaluation phase, only the fine-tuned recognized language model (RLM) is used. The experimental results show that the proposed scheme is effective at classifying intents in the spoken dialogue system containing ASR errors.

Impact of Meeting Early Hearing Detection and Intervention Benchmarks on Spoken Language.

This study compared preschool spoken language outcomes for children with hearing loss who met Early Hearing Detection and Intervention (EHDI) guidelines to those who did not, as well as to compare outcomes for those who met current EHDI guidelines to those who met earlier benchmarks. Finally, the predictive role of meeting each component of the guidelines was evaluated relative to language outcomes. Children who met the EHDI guidelines had higher language scores than those who did not; however, there was no difference between children who met the current guidelines and those who met the earlier benchmarks. Entering early intervention by six months of age was the only unique predictor of spoken language outcomes. The findings suggest that EHDI programs should target increasing the number of children with hearing loss who meet the current 1-3-6 benchmarks with a particular focus on enrollment in early intervention by six months.

Does the cerebellar sequential theory explain spoken language impairments? A literature review.

During the past decades, converging evidence from clinical, neuroimaging and neuroanatomical studies has demonstrated the key role of the cerebellum in the processing of non-motor aspects of language. Although more is known about the way in which the cerebellum participates in the mechanisms involved in written language, there is ambiguous information on its role in other aspects of language, such as in non-motor aspects of spoken language. Thus, to contribute additional insight into this important issue, in the present work, we review several original scientific papers focusing on the most frequent non-motor spoken language impairments evidenced in patients affected by cerebellar pathology, namely, verbal working memory, grammar processing and verbal fluency impairments. Starting from the collected data, we provide a common interpretation of the spoken language disorders in cerebellar patients, suggesting that sequential processing could be the main mechanism by which the cerebellum participates in these abilities. Indeed, according to the cerebellar sequential theory, spoken language impairments could be due to altered cerebellar function to supervise, synchronize and coordinate the activity of different functional modules, affecting the correct optimization of linguistic processing.

No effects of modality in development of locative expressions of space in signing and speaking children.

Linguistic expressions of locative spatial relations in sign languages are mostly visually motivated representations of space involving mapping of entities and spatial relations between them onto the hands and the signing space. These are also morphologically complex forms. It is debated whether modality-specific aspects of spatial expressions modulate spatial language development differently in signing compared to speaking children. In a picture description task, we compared the use of locative expressions for containment, support, and occlusion relations by deaf children acquiring Turkish Sign Language and hearing children acquiring Turkish (age 3;5-9;11). Unlike previous reports suggesting a boosting effect of iconicity, and/or a hindering effect of morphological complexity of the locative forms in sign languages, our results show similar developmental patterns for signing and speaking children's acquisition of these forms. Our results suggest the primacy of cognitive development guiding the acquisition of locative expressions by speaking and signing children.

Adolescents' production of complex syntax in spoken and written expository and persuasive genres.

Analysing spoken and written language samples across different genres provides speech-language pathologists (SLPs) and educators with information about adolescents' production of complex syntax, an important socially and academically related language skill. However, researchers report that production of complex syntax is affected by genre and modality. Although the narrative and expository genres elicit a greater amount of complex syntax than conversational discourse, it is unknown whether differences in production of complex syntax exist between the persuasive and expository genres. The purpose of this study was to compare adolescents' production of complex syntax across spoken and written expository and persuasive genres. Spoken and written expository and persuasive language samples were elicited from 64 adolescents. Complex syntax was measured by calculating per cent of complex utterances and clausal density. Two repeated measures ANOVA revealed that complex syntax production was affected by genre and modality. Adolescents produced a higher percent of complex utterances and a higher clausal density in the persuasive genre than the expository genre. Adolescents produced a higher percent of complex utterances in the written modality than the spoken modality across genres; however, there was not a significant difference in adolescents' clausal density across modalities. There were significant interaction effects between genre and modality for percent of complex utterances and clausal density. The significant interaction effects suggest that differences in the production of complex syntax between the spoken and written modalities depended on the genre elicited. We discuss multiple implications and specific directions for future research.

When it is apt to adapt: Flexible reasoning guides children's use of talker identity and disfluency cues.

An eye-tracking methodology was used to examine whether children flexibly engage two voice-based cues, talker identity and disfluency, during language processing. Across two experiments, 5-year-olds (N = 58) were introduced to two characters with distinct color preferences. These characters then used fluent or disfluent instructions to refer to an object in a display containing items bearing either talker-preferred or talker-dispreferred colors. As the utterance began to unfold, the 5-year-olds anticipated that talkers would refer to talker-preferred objects. When children then encountered a disfluency in the unfolding description, they reduced their expectation that a talker was about to refer to a preferred object. The talker preference-related predictions, but not the disfluency-related predictions, were attenuated during the second half of the experiment as evidence accrued that talkers referred to dispreferred objects with equal frequency. In Experiment 2, the equivocal nature of talkers' referencing was made more apparent by removing neutral filler trials, where objects' colors were not associated with talker preferences. In this case, children ceased making all talker-related predictions during the latter half of the experiment. Taken together, the results provide insights into children's use of talker-specific cues and demonstrate that flexible and adaptive forms of reasoning account for the ways in which children draw on paralinguistic information during real-time processing.

Social inclusion for children with hearing loss in listening and spoken Language early intervention: an exploratory study.

BACKGROUND: Social inclusion is a common focus of listening and spoken language (LSL) early intervention for children with hearing loss. This exploratory study compared the social inclusion of young children with hearing loss educated using a listening and spoken language approach with population data. METHODS: A framework for understanding the scope of social inclusion is presented in the Background. This framework guided the use of a shortened, modified version of the Longitudinal Study of Australian Children (LSAC) to measure two of the five facets of social inclusion ('education' and 'interacting with society and fulfilling social goals'). The survey was completed by parents of children with hearing loss aged 4-5 years who were educated using a LSL approach (n = 78; 37% who responded). These responses were compared to those obtained for typical hearing children in the LSAC dataset (n = 3265). RESULTS: Analyses revealed that most children with hearing loss had comparable outcomes to those with typical hearing on the 'education' and 'interacting with society and fulfilling social roles' facets of social inclusion. CONCLUSIONS: These exploratory findings are positive and warrant further investigation across all five facets of the framework to identify which factors influence social inclusion.

Delayed Anticipatory Spoken Language Processing in Adults with Dyslexia­Evidence from Eye-tracking.

It is now well established that anticipation of upcoming input is a key characteristic of spoken language comprehension. It has also frequently been observed that literacy influences spoken language processing. Here, we investigated whether anticipatory spoken language processing is related to individuals' word reading abilities. Dutch adults with dyslexia and a control group participated in two eye-tracking experiments. Experiment 1 was conducted to assess whether adults with dyslexia show the typical language-mediated eye gaze patterns. Eye movements of both adults with and without dyslexia closely replicated earlier research: spoken language is used to direct attention to relevant objects in the environment in a closely time-locked manner. In Experiment 2, participants received instructions (e.g., 'Kijk naar de(COM) afgebeelde piano(COM)', look at the displayed piano) while viewing four objects. Articles (Dutch 'het' or 'de') were gender marked such that the article agreed in gender only with the target, and thus, participants could use gender information from the article to predict the target object. The adults with dyslexia anticipated the target objects but much later than the controls. Moreover, participants' word reading scores correlated positively with their anticipatory eye movements. We conclude by discussing the mechanisms by which reading abilities may influence predictive language processing.

Spoken language comprehension of phrases, simple and compound-active sentences in non-speaking children with severe cerebral palsy.

BACKGROUND: Children with severe cerebral palsy (CP) (i.e. 'non-speaking children with severely limited mobility') are restricted in many domains that are important to the acquisition of language. AIMS: To investigate comprehension of spoken language on sentence type level in non-speaking children with severe CP. METHODS & PROCEDURES: From an original sample of 87 non-speaking children with severe CP, 68 passed the pre-test (i.e. they matched at least five spoken words to the corresponding objects) of a specifically developed computer-based instrument for low motor language testing (C-BiLLT), admitting them to the actual C-BiLLT computer test. As a result, the present study included 68 children with severe CP (35 boys, 33 girls; mean age 6;11 years, SD 3;0 years; age range 1;9-11;11 years) who were investigated with the C-BiLLT for comprehension of different sentence types: phrases, simple active sentences (with one or two arguments) and compound sentences. The C-BiLLT provides norm data of typically developing (TD) children (1;6-6;6 years). Binomial logistic regression analyses were used to compare the percentage correct of each sentence type in children with severe CP with that in TD children (subdivided into age groups) and to compare percentage correct within the CP subtypes. OUTCOMES & RESULTS: Sentence comprehension in non-speaking children with severe CP followed the developmental trajectory of TD children, but at a much slower rate; nevertheless, they were still developing up to at least age 12 years. Delays in sentence type comprehension increased with sentence complexity and showed a large variability between individual children and between subtypes of CP. Comprehension of simple and syntactically more complex sentences were significantly better in children with dyskinetic CP than in children with spastic CP. Of the children with dyskinetic CP, 10-13% showed comprehension of simple and compound sentences within the percentage correct of TD children, as opposed to none of the children with spastic CP. CONCLUSION & IMPLICATIONS: In non-speaking children with severe CP sentence comprehension is delayed rather than deviant. Results indicate the importance of following comprehension skills across all age groups, even beyond age 12 years. Moreover, the subtype of CP should be considered when establishing an educational programme for sentence comprehension, and augmentative and alternative communication support. In addition, educational programmes for children with severe CP should take into account the linguistic hierarchy of sentence comprehension when focusing on the input and understanding of spoken language comprehension.

A Preliminary Investigation into Effects of Linguistic Abstraction on the Perception of Gender in Spoken Language.

We investigated the role that linguistic abstraction may play in people's perceptions of gender in spoken language. In the first experiment, participants told stories about their best friend and romantic partner. Variations in linguistic abstraction and gender-linked adjectives for describing their close others were examined. Participants used significantly more abstract language to describe men compared to women, possibly reflecting a gender stereotype associated with the dispositionality factor of linguistic abstraction. In a second experiment, a new group of participants judged the gender of the protagonists from the stories generated in Experiment 1, after the explicit linguistic gender cues were removed. Consistent with the dispositionality factor, linguistic abstraction moderated the effects of the gender stereotypicality of the context (masculine, feminine, or neutral) on participants' gender judgments. Discussion focuses on the implications of the results for the communication of gender stereotypes and the effects of linguistic abstraction in more naturalistic language.

Automatic selection of spoken language biomarkers for dementia detection.

This paper analyzes diverse features extracted from spoken language to select the most discriminative ones for dementia detection. We present a two-step feature selection (FS) approach: Step 1 utilizes filter methods to pre-screen features, and Step 2 uses a novel feature ranking (FR) method, referred to as dual dropout ranking (DDR), to rank the screened features and select spoken language biomarkers. The proposed DDR is based on a dual-net architecture that separates FS and dementia detection into two neural networks (namely, the operator and selector). The operator is trained on features obtained from the selector to reduce classification or regression loss. The selector is optimized to predict the operator's performance based on automatic regularization. Results show that the approach significantly reduces feature dimensionality while identifying small feature subsets that achieve comparable or superior performance compared with the full, default feature set. The Python codes are available at https://github.com/kexquan/dual-dropout-ranking.