Discriminative Learning and the Lexicon: NDL and LDL

Naive discriminative learning (NDL) and linear discriminative learning (LDL) are simple computational algorithms for lexical learning and lexical processing. Both NDL and LDL assume that learning is discriminative, driven by prediction error, and that it is this error that calibrates the association strength between input and output representations. Both words’ forms and their meanings are represented by numeric vectors, and mappings between forms and meanings are set up. For comprehension, form vectors predict meaning vectors. For production, meaning vectors map onto form vectors. These mappings can be learned incrementally, approximating how children learn the words of their language. Alternatively, optimal mappings representing the end state of learning can be estimated. The NDL and LDL algorithms are incorporated in a computational theory of the mental lexicon, the ‘discriminative lexicon’. The model shows good performance both with respect to production and comprehension accuracy, and for predicting aspects of lexical processing, including morphological processing, across a wide range of experiments. Since, mathematically, NDL and LDL implement multivariate multiple regression, the ‘discriminative lexicon’ provides a cognitively motivated statistical modeling approach to lexical processing.

The minor T allele of the MUC5B promoter rs35705950 associated with susceptibility to idiopathic pulmonary fibrosis: a meta-analysis

MUC5B promoter rs35705950 T/G gene polymorphism has been associated with the risk of IPF, but the influence of this relationship varies among different populations. In the past 2 years, there were new clinical studies with different results, but none of them reached unified conclusions. Therefore, this study further included the latest case–control studies, integrated their results and carried out meta-analysis on them to draw reliable conclusions. PubMed, EMBASE, CNKI, Wanfang database and VIP Chinese science were searched by a computer to collect the related literatures of MUC5B gene polymorphism and IPF susceptibility published before June 15, 2021. The first author, year of publication, diagnostic criteria and gene frequency were extracted after screened them. Forest plot was drawn and the trial sequential analysis (TSA) was carried out to confirm the stability of the meta-analysis results. Registration number: CRD42021272940. A total of 24 case–control studies (13 studies on the Caucasian, 7 studies on the Asian and 4 studies on the mixed population), and a total of 6749 IPF patients and 13,898 healthy controls were included in this study. The T vs.G, TT vs. GG, GT vs. GG, GT + TT vs. GG and TT vs. GG + GT genetic models of MUC5B promoter rs35705950 T/G polymorphism were associated with IPF risk in all populations, and the effect values were ([OR] 4.12, 95% CI [3.64, 4.67]), ([OR] 10.12, 95% CI [7.06, 14.49]), ([OR] 4.84, 95% CI [3.85, 6.08]), ([OR] 4.84, 95% CI [3.79, 6.19]) and ([OR] 5.11, 95% CI [4.02, 6.49]), respectively. The results of TSA confirmed the stability of the results. Subgroup analysis showed that T vs.G, TT vs. GG, GT vs. GG, GT + TT vs. GG and TT vs. GG + GT genetic models of MUC5B polymorphism were associated with IPF risk in Caucasian population. The effect values were ([OR] 4.50, 95% CI [3.93, 5.16]), ([OR] 10.98, 95% CI [7.59, 15.89]), ([OR] 6.27, 95% CI [5.37, 7.32]), ([OR] 6.30, 95% CI [5.19, 7.64]) and ([OR] 5.15, 95% CI [4.01, 6.61]), respectively. Similar results were also found in Asian and mixed populations. The association strength of the minor T allele in the Caucasian was more significant than that of the Asian population ([OR] 4.50 vs. [OR] 2.39), and the association strength of all genetic models carrying "T" was more significant than that of the Asian population ([OR] 10.98 vs. [OR] 4.29). In Caucasian, Asian and mixed populations, T minor allele carriers were more likely to be susceptible to pulmonary fibrosis, and TT genotype carriers were more likely to be susceptible to IPF than GT genotype carriers. The association between IPF and Caucasian population with minor T allele and all "T" genetic model was more significant than that of Asian population.

Vocabulary development through reading: A comparison of approaches

<p>This thesis compares two approaches to extensive reading to determine the extent that they facilitate vocabulary development. The first approach is a traditional reading-only approach, and the second approach is a task-based approach which supplements reading with post-reading meaning-focused discussions. These two approaches are compared using a battery of tests, most notably a measure for productive knowledge of word associations. For years, scholars have believed that word associations have potential to reveal important information about a person’s language proficiency. One reason word associations are intriguing is that a large amount of a person’s lexicon can be assessed (Meara, 2009). This is possible because a large amount of data from the learner can be gathered in a short period of time. Another intriguing aspect of word association data is that it is one aspect of vocabulary knowledge that is not based on correct performance. This raises the question of an appropriate means of assigning value to the associations, a question which still hinders research to this day. Recent research has made progress in this area with a multi-level taxonomy (i.e., Fitzpatrick, 2007), creating a picture of the types of associations which exist in a learner’s lexicon. However, this taxonomy does not address the strength of the association. Wilks and Meara (2007) have attempted to tackle association strength through the use of self-report measures, whereby a test-taker reports strength of association on a four-point scale from weak to strong. This has left them with "...problems which we have not yet solved, notably a tendency for some test takers to claim that most associations are strong, while others appear to be very reluctant to identify strong associations..." (Meara, 2009, p. 80). In other words, the question of how to appropriately determine association strength is still unanswered. In the current study lexical development, in the form of word association knowledge, was measured using a multi-response word association test. Participants were assessed on their knowledge of 60 target words which occurred in five graded readers that they read over the course of the study. The learners first self-reported their knowledge of the 60 target words in terms of no knowledge, form knowledge, or meaning knowledge. The students provided up to five associations for each word that they reported at either the form or meaning levels. They did this once before reading the five graded readers, and again after finishing the graded readers. The associations provided by the students were analyzed using Latent Semantic Analysis, a method for computing semantic similarity between words (Landauer & Dumais, 1997). The associations a learner provided for each target word were assigned a similarity value representing how similar they were to the target word to which they were provided. The hypothesis was that the students who engaged in the post-reading discussion activities would show greater increases in associational knowledge of the target words than those students who did not participate in the discussions. The major finding from this thesis was that the students who struggled with a word during the post-reading discussion and were provided an opportunity to discuss the word with their group developed associational knowledge to a significantly greater degree than those students who did not encounter the words during the discussions. This emphasizes the facilitative role that meaning-focused output activities have on vocabulary development. In addition, the associational knowledge developed at the initial stages of word learning (i.e., from no knowledge to form knowledge), continued to develop from form knowledge of a word to meaning knowledge of the word, and was also developing even when words did not change in reported knowledge. This suggests a continual restructuring of the learners’ lexicon, exemplifying past research (e.g., Henriksen, 1999). Overall, the findings suggest that an extensive reading approach which includes opportunities for meaning-focused interaction has greater benefits for lexical development when compared to a traditional reading-only approach to extensive reading.</p>

Vocabulary development through reading: A comparison of approaches

<p>This thesis compares two approaches to extensive reading to determine the extent that they facilitate vocabulary development. The first approach is a traditional reading-only approach, and the second approach is a task-based approach which supplements reading with post-reading meaning-focused discussions. These two approaches are compared using a battery of tests, most notably a measure for productive knowledge of word associations. For years, scholars have believed that word associations have potential to reveal important information about a person’s language proficiency. One reason word associations are intriguing is that a large amount of a person’s lexicon can be assessed (Meara, 2009). This is possible because a large amount of data from the learner can be gathered in a short period of time. Another intriguing aspect of word association data is that it is one aspect of vocabulary knowledge that is not based on correct performance. This raises the question of an appropriate means of assigning value to the associations, a question which still hinders research to this day. Recent research has made progress in this area with a multi-level taxonomy (i.e., Fitzpatrick, 2007), creating a picture of the types of associations which exist in a learner’s lexicon. However, this taxonomy does not address the strength of the association. Wilks and Meara (2007) have attempted to tackle association strength through the use of self-report measures, whereby a test-taker reports strength of association on a four-point scale from weak to strong. This has left them with "...problems which we have not yet solved, notably a tendency for some test takers to claim that most associations are strong, while others appear to be very reluctant to identify strong associations..." (Meara, 2009, p. 80). In other words, the question of how to appropriately determine association strength is still unanswered. In the current study lexical development, in the form of word association knowledge, was measured using a multi-response word association test. Participants were assessed on their knowledge of 60 target words which occurred in five graded readers that they read over the course of the study. The learners first self-reported their knowledge of the 60 target words in terms of no knowledge, form knowledge, or meaning knowledge. The students provided up to five associations for each word that they reported at either the form or meaning levels. They did this once before reading the five graded readers, and again after finishing the graded readers. The associations provided by the students were analyzed using Latent Semantic Analysis, a method for computing semantic similarity between words (Landauer & Dumais, 1997). The associations a learner provided for each target word were assigned a similarity value representing how similar they were to the target word to which they were provided. The hypothesis was that the students who engaged in the post-reading discussion activities would show greater increases in associational knowledge of the target words than those students who did not participate in the discussions. The major finding from this thesis was that the students who struggled with a word during the post-reading discussion and were provided an opportunity to discuss the word with their group developed associational knowledge to a significantly greater degree than those students who did not encounter the words during the discussions. This emphasizes the facilitative role that meaning-focused output activities have on vocabulary development. In addition, the associational knowledge developed at the initial stages of word learning (i.e., from no knowledge to form knowledge), continued to develop from form knowledge of a word to meaning knowledge of the word, and was also developing even when words did not change in reported knowledge. This suggests a continual restructuring of the learners’ lexicon, exemplifying past research (e.g., Henriksen, 1999). Overall, the findings suggest that an extensive reading approach which includes opportunities for meaning-focused interaction has greater benefits for lexical development when compared to a traditional reading-only approach to extensive reading.</p>

N-gram use in EFL learners’ retelling and monologic tasks

This study examined n-gram use in oral production by Chinese college-level English as a foreign language (EFL) learners at four distinct proficiency levels. Thirty indices regarding range, frequency, and association strength of bi- and tri-grams obtained from retelling and monologic samples were analyzed. Results suggest that, i) the four proficiency levels differed in measures for frequency and association strength of bi- and tri-grams, ii) academic bi- and tri-gram proportions and association strength (captured by MI- and t-scores) were predictive of EFL speaking proficiency for both the retelling and monologic samples but the effects were small, and iii) EFL learners used more well-attested bi- and tri-grams in monologues than in retelling, demonstrating that higher rated samples tended to contain more strongly-associated bi- and tri-grams, a greater proportion of frequent attested academic tri-grams, and that EFL n-gram use was task-sensitive. These findings help enrich our understanding on EFL development of multi-word sequences and have potentially useful implications for EFL pedagogy.

The Challenge of Choosing the Best Classification Method in Radiomic Analyses: Recommendations and Applications to Lung Cancer CT Images

Radiomics uses high-dimensional sets of imaging features to predict biological characteristics of tumors and clinical outcomes. The choice of the algorithm used to analyze radiomic features and perform predictions has a high impact on the results, thus the identification of adequate machine learning methods for radiomic applications is crucial. In this study we aim to identify suitable approaches of analysis for radiomic-based binary predictions, according to sample size, outcome balancing and the features–outcome association strength. Simulated data were obtained reproducing the correlation structure among 168 radiomic features extracted from Computed Tomography images of 270 Non-Small-Cell Lung Cancer (NSCLC) patients and the associated to lymph node status. Performances of six classifiers combined with six feature selection (FS) methods were assessed on the simulated data using AUC (Area Under the Receiver Operating Characteristics Curves), sensitivity, and specificity. For all the FS methods and regardless of the association strength, the tree-based classifiers Random Forest and Extreme Gradient Boosting obtained good performances (AUC ≥ 0.73), showing the best trade-off between sensitivity and specificity. On small samples, performances were generally lower than in large–medium samples and with larger variations. FS methods generally did not improve performances. Thus, in radiomic studies, we suggest evaluating the choice of FS and classifiers, considering specific sample size, balancing, and association strength.

P168 Depression and fatigue dependent variation of gut microbiome in patients with active Inflammatory Bowel Disease

Background Fatigue and depression are highly prevalent extraintestinal symptoms in IBD, especially during active disease. The underlying mechanisms are still poorly understood, but an involvement of the microbiota-gut-brain-axis seems likely. The aim of this study was to examine associations between the gut microbiota and symptoms of fatigue and depression in patients with active IBD. Methods We included 62 patients with active IBD. Blood and stool samples were collected before a change of therapy and patients completed questionnaires regarding fatigue (WEIMuS) and depression (HADS). Fecal microbiota were analyzed by means of shotgun metagenomic sequencing. Based on taxonomical and functional metagenomic annotations of gut microbiome, we investigated the correlation between functional modules and psychometric scores as well as the nature of co-occurrence networks on the genus level using Bayesian statistical methods and analyzed modulating effects of fatigue and depression on genera co-occurrence. CRP levels as well as age, gender, diagnosis and medication were included as nuisance variables. Results Half of the patients (n=31) were fatigued (WEIMuS &gt;/= 32P.), and 18 patients reported symptoms of at least mild to moderate depression (HADS D &gt;/=10P.). Co-occurrence of genera that is influenced by depression or fatigue in this active IBD-sample with is displayed in Figure 1 (Log10(BF10) ≥ 1.0, strong evidence for H1). Evidence for an influence of depression severity was found for 10 genera combinations, with the strongest effects on the Dorea-Butyrococcus-association (Log10(BF10) = 1.41, increase of association strength with higher depression score) and the association between an unknown genus of the Coriobacteriaceae family with Arabia (Log10(BF10) = 1.00, decrease of association strength with higher depression score). Modulatory effects of fatigue severity was present in 10 genera combinations, most pronounced for the Azospirillum-Akkermansia- (Log10(BF10) =2.39, increase of association strength with higher fatigue-score) and Odoribacter-Actinomyces-association (Log10(BF10) =1.58, decrease of association strength with higher fatigue-score) (Figure 2). Conclusion This study is the first to address the relationship between extraintestinal symptoms like fatigue and depression and microbiota parameters in patients with active IBD by means of network analysis with Bayesian methods. Our results show strong associations between these symptoms and co-occurring genera in a sample-specific network, indicating a role of the microbiota-gut-brain-axis in the development of fatigue and depression during active disease.

Assessing the efficiency of COVID-19 NPIs in France: a retrospective study using a novel methodology

After an initial phase of low reactivity from the French public health authorities, in the face of the emergence of SARS-CoV-2 in February 2020, various Non Pharmaceutical Interventions (NPIs) were put in place (strict stay-at-home orders, followed by mandatory mask wearing in public places, curfews, partial lockdowns, etc.). In our knowledge, no study has independently assessed their respective effectiveness in an independent manner nor a synergistic manner. Our study has retrospectively studied (from 03/01/2020 to 30/01/2021), using metropolitan France data, the association strength (using normalized mutual information) as well as the linear correlation (using Pearson’s correlation) between more restrictive NPIs (mrNPIs) and epidemiological markers of COVID-19. All mrNPIs were moderately associated with a viral reproduction rate decrease but were associated neither with a decrease in COVID-19 daily hospitalizations, nor with COVID-19 daily ICU admissions. This paper is only for academic discussion, conclusions need to be confirmed by further research. Data and codes were available here http://gitlab.com/covid-data-2/lockdown-and-curfew.

Improvement on the association strength: implementing a probabilistic measure inspired on combinations without repetition.

The use of co-occurrence data is common in various domains. Co-occurrence data often needs to be normalised to correct for the size-effect. To this end, van Eck and Waltman (2009) recommend a probabilistic measure known as the association strength. However, this formula, based on combinations with repetition, implicitly assumes that observations from the same entity can co-occur even though in the intended usage of the measure these self-co-occurrences are non-existent. A more accurate measure inspired on combinations without repetition is introduced here and compared to the original formula in mathematical derivations, simulations, and patent data, which shows that the original formula overestimates the relation between a pair and that some pairs are more overestimated than others. The new measure is available in the EconGeo package for R maintained by Balland (2016). Peer Review https://publons.com/publon/10.1162/qss_a_00122