The Diversity of Classifier Inventory in Mandarin Dialects: A Case Study of Baoding

Our study compares Standard Mandarin (the Beijing dialect used in spoken and written registers) with the Mandarin dialect of Baoding (one of the Mandarin dialects belonging to the Jì-lŭ Mandarin group, Hebei-Shandong). Standard Mandarin and Baoding are geographically and phylogenetically closely related, but they differ in terms of their classifier system, as Standard Mandarin resorts to a wide array of sortal classifiers whereas Baoding only uses one general classifier. We first provide a detailed analysis of the unconventional classifier system in Baoding. Then, we compare the lexical and discourse functions of sortal classifiers in Standard Mandarin and Baoding. We show that Standard Mandarin does present a certain level of convergence with its geographical neighbour Baoding. However, these varieties also display significant divergences, as several lexical and discourse functions typically associated with classifier systems cannot be fulfilled by the only classifier found in Boading.

The Semantics of Chinese Noun Phrases

Chinese nominal phrases are typologically distinct from their English counterparts in many aspects. Most strikingly, Chinese is featured with a general classifier system, which not only helps to categorize nouns but also has to do with the issue of quantification. Moreover, it has neither noncontroversial plural markers nor (in)definite markers. Its bare nouns are allowed in various argument positions. As a consequence, Chinese is sometimes characterized as a classifier language, as an argumental language, or as an article-less language. One of the questions arising is whether these apparently different but related properties underscore a single issue: that it is the semantics of nouns that is responsible for all these peculiarities of Mandarin nominal phrases. It has been claimed that Chinese nouns are born as kind terms, from which the object-level readings can be derived, being either existential or definite. Nevertheless, the existence of classifiers in Chinese is claimed to be independent of the kind denotation of its bare nouns. Within the general area of noun semantics, a number of other semantic issues have generated much interest. One is concerned with the availability of the mass/count distinction in Mandarin nominal phrases. Another issue has to do with the semantics of classifiers. Are classifiers required by the noun semantics or the numeral semantics, when occurring in the syntactic context of Numeral/Quantifier-Classifier-Noun? Finally, how is the semantic notion of definiteness understood in article-less languages like Mandarin Chinese? Should its denotation be characterized with uniqueness or familiarity?

A Novel Unsupervised Computational Method for Ventricular and Supraventricular Origin Beats Classification

Arrhythmias are the most common events tracked by a physician. The need for continuous monitoring of such events in the ECG has opened the opportunity for automatic detection. Intra- and inter-patient paradigms are the two approaches currently followed by the scientific community. The intra-patient approach seems to resolve the problem with a high classification percentage but requires a physician to label key samples. The inter-patient makes use of historic data of different patients to build a general classifier, but the inherent variability in the ECG’s signal among patients leads to lower classification percentages compared to the intra-patient approach. In this work, we propose a new unsupervised algorithm that adapts to every patient using the heart rate and morphological features of the ECG beats to classify beats between supraventricular origin and ventricular origin. The results of our work in terms of F-score are 0.88, 0.89, and 0.93 for the ventricular origin beats for three popular ECG databases, and around 0.99 for the supraventricular origin for the same databases, comparable to supervised approaches presented in other works. This paper presents a new path to make use of ECG data to classify heartbeats without the assistance of a physician despite the needed improvements.

Implications on Nominal Argument Formation in General

Chapter 6 discusses the relationship between argument formation in classifier languages and argument formation more generally. It begins with a discussion on the variable and uniform properties concerning nominal arguments in NMLs and shows that their variation can be captured by two of the parameters in classifier languages proposed in Chapter 5. It further shows that the variation in whether nouns are coded as kinds or properties can differentiate classifier languages from NMLs. It then discusses languages, which have neither a general classifier system (unlike Mandarin or Nuosu Yi) nor obligatory singular/plural marking on nouns (unlike English or Hindi). Building on previous analyses, this chapter analyzes Yudja as a language with a silent Cl and Lhasa Tibetan a language with a silent Div. This chapter allows us to further update the variation and typology of nominal argument formation and to predict more types of languages.

USING NEURAL NETWORK TO UNCOVER CELL TYPES THROUGH SNRNA SEQUENCING

The advent of single-nucleus RNA-sequencing (snRNAseq) has allowed for the exploration of genetic signatures of the numerous cells in the brain. In particular, snRNAseq data can provide new insights into how many neurodegenerative diseases, such as Alzheimer’s Disease, alter cells in the brain. One major challenge with analyzing snRNAseq data is the lack of a systematic way to classify the various cell types across different datasets. To address this challenge, we developed a general classifier (“DeepSeq”) that uses state-of-the-art deep learning approaches. We trained our model on multiple snRNAseq datasets derived from post-mortem brain tissue in individuals with and without clinical diagnosis of Alzheimer’s Disease from the ROSMAP cohorts. The two snRNAseq datasets contained 70,064 nuclei and 170,275 nuclei. The two studies employed different clustering techniques, and identified 44 and 18 putative cell types. To map these disparate cluster identities across datasets, we extracted the most relevant genes and trained two separate networks, one on each dataset. We then validated each classifier separately on the holdout cells. The resulting classifier accuracy were 87% and 94%. To map clusters across datasets, we then applied each classifier to the other dataset. Both classifiers yielded mappings that reflected the overall biology, correctly categorizing the nuclei into broad and fine cell type classes. Although validation on additional datasets would expand the generality of this approach, our results show that DeepSeq is an easily implementable classification tool that can assign identity to nuclei in new snRNAseq datasets without the need for preprocessing or cross-batch alignment.

The coalescence of grammatical gender and numeral classifiers in the general classifier wota in Nepali

While nominal classification has received considerable attention, relatively little is known about cross-linguistically rare complex systems. An example is provided by Nepali (Indo-European, Indic), which possesses both grammatical gender and numeral classifiers. Our aim is to examine morphosyntactic and functional properties of the general classifier wota. Unusually, the classifier exhibits gender agreement both in its independent forms and as fused with a numeral, raising questions about its lexical and pragmatic functions. Our study contributes to the typology of nominal classification by proposing a functional approach to cases of complex co-occurrence of gender and classifiers.

Combining Lexical and Syntactic Features for Detecting Content-Dense Texts in News

Content-dense news report important factual information about an event in direct, succinct manner. Information seeking applications such as information extraction, question answering and summarization normally assume all text they deal with is content-dense. Here we empirically test this assumption on news articles from the business, U.S. international relations, sports and science journalism domains. Our findings clearly indicate that about half of the news texts in our study are in fact not content-dense and motivate the development of a supervised content-density detector. We heuristically label a large training corpus for the task and train a two-layer classifying model based on lexical and unlexicalized syntactic features. On manually annotated data, we compare the performance of domain-specific classifiers, trained on data only from a given news domain and a general classifier in which data from all four domains is pooled together. Our annotation and prediction experiments demonstrate that the concept of content density varies depending on the domain and that naive annotators provide judgement biased toward the stereotypical domain label. Domain-specific classifiers are more accurate for domains in which content-dense texts are typically fewer. Domain independent classifiers reproduce better naive crowdsourced judgements. Classification prediction is high across all conditions, around 80%.

Implementation of Evolution Strategies for Classifier Model Optimization

<span style="font-size: 9.0pt; mso-bidi-font-size: 11.0pt; line-height: 107%; font-family: 'Times New Roman','serif'; mso-fareast-font-family: Calibri; mso-fareast-theme-font: minor-latin; mso-ansi-language: EN-US; mso-fareast-language: EN-US; mso-bidi-language: AR-SA;">Classification becomes one of the classic problems that are often encountered in the field of artificial intelligence and data mining. The problem in classification is how to build a classifier model through training or learning process. Process in building the classifier model can be seen as an optimization problem. Therefore, optimization algorithms can be used as an alternative way to generate the classifier models. In this study, the process of learning is done by utilizing one of Evolutionary Algorithms (EAs), namely Evolution Strategies (ES). Observation and analysis conducted on several parameters that influence the ES, as well as how far the general classifier model used in this study solve the problem. The experiments and analyze results show that ES is pretty good in optimizing the linear classification model used. For Fisher’s Iris dataset, as the easiest to be classified, the test accuracy is best achieved by 94.4%; KK Selection dataset is 84%; and for SMK Major Election datasets which is the hardest to be classified reach only 49.2%.</span>