Search

Sentiment analysis is helpful to bestow ability of understanding human’s attitude in texts on artificial intelligence systems. In this area, text sentiment is usually signaled by a few indicative words that convey affective meanings and arouse readers’ collective emotions. However, most existing sentiment analysis models have predominantly featured through neural network architectures with end-to-end training manner and limited awareness of affective knowledge, which, as a result, often fails to pinpoint the essential features for sentiment prediction. In this work, we present a novel approach for sentiment analysis by fusing external affective knowledge into neural networks. The affective knowledge is distilled from two sentiment lexicons grounded by two psychological theories, e.g., the Affect Control Theory and word affections in terms of Valence, Arousal, and Dominance. To examine the effects of affective knowledge over sentiment analysis, we conduct cross-dataset and cross-model experiments along with a detailed ablation analysis. Results show that our proposed method outperforms trendy neural networks in all the five benchmarks with consistent and significant improvement (1.4% Accuracy in average). Further discussions demonstrate that all affective attributes exhibit positive effects to model enhancement and our model is robust to the change of lexicon size.

This paper investigates the emergence of COVID-19 neologisms. It focuses on the strategies used to coin emerging neologisms, the relationship between the strategies and the usage preferences, as well as the correlation between internet usage data and epidemiological data. The internet usage data were collected from December 2019 to June 2020 from the Baidu Index, covering the usage of all five categories of the COVID-19 name variants. The epidemiological data, from the Chinese Center for Disease Control and Prevention, are statistics of newly confirmed cases, newly suspected cases, new deaths, and currently suspected cases at a given time. The study identified three strategies in the coinage of neologisms: categorization, avoidance, and synthesis. In addition, a strong correlation between emergent neologisms and pandemic developments was discovered with a binomial model, and the emerging neologisms demonstrated a skewed S-curve life cycle, which is different from the established S-curve model of replacement changes. In sum, by leveraging internet usage data, this first study of the life cycle of emergent neologisms has several contributions: A theory of how new words emerge, the correlation between emergent neologisms and emerging events, and the potential of modeling language use for epidemiological predictions.

In this article, we expand the typological studies on weather expressions by bridging linguistic and meteorological ontologies. Based on our investigations into weather words of Sinitic languages from both synchronic and diachronic perspectives, we propose a new weather event typology, typology of meteorological events (TyME), with two binary features, [±Process] and [±Material]. We argue that this typology covers more weather phenomena in a systematic and ontologically transparent way and can benefit synchronic and diachronic studies on weather and language. In addition, a cross-linguistic investigation is conducted on previously less studied meteorological expressions: fog, dew and frost. The results show that fog, dew and frost can be said to fall in the majority of the languages, which seems to contradict their meteorological formation behaviours, but in fact conforms to natural laws. Based on the new weather event typology and analysed data, we discover that fog, dew and frost all correlate with precipitation in terms of directionality and encoding types. The two binary features we propose account for these formerly overlooked weather events as well as others and can provide effective assistance in analysing the mechanisms underlying those seemingly scientifically infelicitous expressions.

This paper proposes a QP account for indefiniteness in a range of languages, with QP being situated between DP and NumP. This proposal differs from the traditional DP framework crucially in two aspects: (i) definiteness and indefiniteness are expressed by DP and QP, respectively, and their co-occurrence is theoretically allowed, under due circumstances; (ii) both [Spec, DP] and [Spec, QP] can serve as the landing sites for an XP under a wh-movement, which can thus be associated with both definiteness and indefiniteness. Crucially, the postulation of QP accounts for several nominal phenomena in the Sinhala language, which pose serious challenges to the traditional DP framework. First, Sinhala features the existence of indefinite articles as well as the absence of definite ones; second, an indefinite article can co-occur with a demonstrative in one and the same nominal phrase; third, NP necessarily precedes [numeral + (classifier)], regardless of its (in-)definiteness; fourth, indefinite quantifiers exhibit disparities along the line of their (in-)sensitivity toward the feature of animacy. All these phenomena can be adequately captured by the proposal of QP; specifically, the first three are explained by the DP–QP division whereas the fourth one by the QP–NumP distinction. The DP proposal is further supported by the need to have both DP–QP and QP–NumP divisions in the Chinese language, as well as the intuitive account it provides for the “double definiteness” in Scandinavian languages. In sum, the QP postulation is strongly motivated by cross-lingual evidence for the account of definiteness and indefiniteness as two separate but related unary features instead of binary value of the same feature, e.g. [+/−DEF].

In sentiment analysis of product reviews, both user and product information are proven to be useful. Current works handle user profile and product information in a unified model which may not be able to learn salient features of users and products effectively. In this work, we propose a dual user and product memory network (DUPMN) model to learn user profiles and product information for reviews classification using separate memory networks. Then, the two representations are used jointly for sentiment analysis. The use of separate models aims to capture user profiles and product information more effectively. Comparing with state-of-the-art unified prediction models, evaluations on three benchmark datasets (IMDB, Yelp13, and Yelp14) show that our dual learning model gives performance gain of 0.6%, 1.2%, and 0.9%, respectively. The improvements are also deemed very significant measured by p-values.

This paper examines the mapping directionality tendencies of linguistic synesthesia in Mandarin using a corpus-based approach. Based on this set of less-studied data, we find that Mandarin synesthesia does not share the same directionality tendencies with linguistic synesthesia in Indo-European languages, which challenges the assumed cross-linguistic universality of these transfer patterns. Based on the corpus data, we demonstrate that there are three types of directional tendencies for Mandarin synesthesia: unidirectional, biased-directional, and bidirectional. Unidirectional synesthesia is rule-based, while synesthesia that is biased in one direction is frequency-based. In contrast, bidirectional synesthesia shows no directional preference. Thus, the directionality of linguistic synesthesia cannot be interpreted as rule-based or frequency-based exclusively. In addition, this study finds that linguistic synesthesia shows language-specific variations for directionality tendencies grounded in both embodiment and neural mechanisms, which challenges the theory that linguistic synesthesia is a bio-neurologically based linguistic realization. Lastly, the fact that linguistic synesthesia involves both rule-based and frequency-based transfer directionalities suggests that the relationship between linguistic synesthesia and metaphor merits further exploration.

Highlights? Features are extracted with various lightweight methods and from broad resources. ? The unsupervised features improve the robustness of a disambiguation system. ? Our AE system integrates various extraction approaches with high precision. ? Each integrated AE approach exactly extracts some of the right target information. Personal information extraction, which extracts the persons in question and their related information (such as biographical information and occupation) from web, is an important component to construct social network (a kind of semantic web). For this practical task, two important issues are to be discussed: personal named entity ambiguity and the extraction of personal information for a specific person. For personal named entity ambiguity, which is a common phenomenon in the fast growing web resource, we propose a robust system which extracts lightweight features with a totally unsupervised approach from broad resources. The experiments show that these lightweight features not only improve the performances, but also increase the robustness of a disambiguation system. To extract the information of the focus person, an integrated system is introduced, which is able to effectively re-use and combine current well-developed tools for web data, and at the same time, to identify the expression properties of web data. We show that our flexible extraction system achieves state-of-the-art performances, especially the high precision, which is very important for real applications.

This paper challenges the traditional view that nominal classifiers classify individuals. Instead, we suggest that classifiers coerce nouns to refer to kinds and events as well as to individuals. This finding argues against the view that nouns refer only to entities, and suggests that classifiers do not simply agree with a noun, but instead coerce a particular meaning from it. Moreover, the Mandarin classifier system creates a taxonomic system involving events, kinds and individuals respectively. Within each classifier type an independent classification system of the collocating noun type is created. These findings are important first because they emphasize that the understanding of the semantics of nouns involves more than simple reference to an individual entity. Second, it is the first time that the previously abstract semantic distinctions among kinds, individuals and events, as well as within kinds and within events, have been found to be instantiated in a particular system of a natural language grammar, namely, the classifier system.