Commemorative Practices in Urban Discourse Space: Historical-Political Projection

The paper focuses on the potential of urban communicative space to transmit basic ideology paradigm and values of the current political regime in the course of interaction with city residents. Commemoration is regarded as one the key entities involved in transmitting ideologically charged information. The phenomenon of commemoration is understood as a collection of public practices aimed at creating government-approved values and behavioral models via regular reproduction schemes implemented on the basis of perception of past recognized in the society. The goal of the research was to examine verbalization of commemoration in urban space with regard to the historical-political context. The empirical basis of the study includes a corpus of commemorative texts collected by the authors. Commemorative texts were extracted from the open data portal developed and supported by the Government of Moscow. The total number of records registered in the corpus amounted to over 1700. The language data were processed via the AntConc software that allows obtaining information about word frequency and the contexts in which the relevant word occurs. In the next step conclusions about topical and conceptual dominants of commemorative texts were made. Further investigation allowed describing the structural scheme of commemorative texts, determining its zero variability in different temporal periods, identifying an obligatory structural element that displayed sensitiveness to political climate and specifying key discourse strategies correlated with the ideological paradigm of the current political regime.

Natural Language Processing through the Subtractive Mountain Clustering Algorithm - A Medication Intake Chatbot

In this work, the subtractive mountain clustering algorithm has been adapted to the problem of natural languages processing in view to construct a chatbot that answers questions posed by the user. The implemented algorithm version allosws for the association of a set of words into clusters. After finding the centre of every cluster — the most relevant word, all the others are aggregated according to a defined metric adapted to the language processing realm. All the relevant stored information (necessary to answer the questions) is processed, as well as the questions, by the algorithm. The correct processing of the text enables the chatbot to produce answers that relate to the posed queries. Since we have in view a chatbot to help elder people with medication, to validate the method, we use the package insert of a drug as the available information and formulate associated questions. Errors in medication intake among elderly people are very common. One of the main causes for this is their loss of ability to retain information. The high amount of medicine intake required by the advanced age is another limiting factor. Thence, the design of an interactive aid system, preferably using natural language, to help the older population with medication is in demand. A chatbot based on a subtractive cluster algorithm is the chosen solution.

Semantic and motivational characteristics of enmity vocabulary: the relevance of structural motivation

The object of the research includes the actual designations of the subject with the semantics of enmity, formed with the active foreign language word-formation components, functioning in Russian-language online media and Internet communication in 2000-2020. The approaches used include system-centric and text-centric, semasiological and onomasiological approaches, motivational, definitional, functional-semantic and contextual analysis. It analyses the syntagmatics, semantics, word's inner form, type of motivation, motivational form and meaning, motivational and classification features, lexical and structural motivators, ways of discursive actualising the motivational relations of the studied words. It is shown that selected lexical units with the component –phobe, -phrenic, - saur, -down, -hater, -oid, -oholik, -path, -man, -(e)rast belong to the vocabulary of enmity depending on their significative or pragmatic component, implement a negative assessment of intellectual, psychological, moral qualities of the subject. The actual vectors of developing the nominal vocabulary of enmity in the Russian language are determined by integration, intensification, internationalization, intensification. The relevant word-forming tendencies in the studied group of nouns are highlighted – the frequency of word composition, non-usual ways of word formation, nominations by analogy, the increasing role of onyms, the activity of word-forming components with a negative rating. It is demonstrated that motivational relations of lexemes are discursively implemented through the actualization of lexical and structural motivation, the paradigmatic value of lexemes, the subjective modality that the addressee uses, his/her individual motivation of words.

Subtractive Mountain Clustering Algorithm Applied to a Chatbot to Assist Elderly People in Medication Intake

Errors in medication intake among elderly people are very common. One of the main causes for this is their loss of ability to retain information. The high amount of medicine intake required by the advanced age is another limiting factor. Thence, the design of an interactive aid system, preferably using natural language, to help the older population with medication is in demand. A chatbot based on a subtractive cluster algorithm, included in unsupervised learned models, is the chosen solution since the processing of natural languages is a necessary step in view to construct a chatbot able to answer questions that older people may pose upon themselves concerning a particular drug. In this work, the subtractive mountain clustering algorithm has been adapted to the problem of natural languages processing. This algorithm version allows for the association of a set of words into clusters. After finding the centre of every cluster — the most relevant word, all the others are aggregated according to a defined metric adapted to the language processing realm. All the relevant stored information is processed, as well as the questions, by the algorithm. The correct processing of the text enables the chatbot to produce answers that relate to the posed queries. To validate the method, we use the package insert of a drug as the available information and formulate associated questions.

Dual-objective fine-tuning of BERT for entity matching

An increasing number of data providers have adopted shared numbering schemes such as GTIN, ISBN, DUNS, or ORCID numbers for identifying entities in the respective domain. This means for data integration that shared identifiers are often available for a subset of the entity descriptions to be integrated while such identifiers are not available for others. The challenge in these settings is to learn a matcher for entity descriptions without identifiers using the entity descriptions containing identifiers as training data. The task can be approached by learning a binary classifier which distinguishes pairs of entity descriptions for the same real-world entity from descriptions of different entities. The task can also be modeled as a multi-class classification problem by learning classifiers for identifying descriptions of individual entities. We present a dual-objective training method for BERT, called JointBERT, which combines binary matching and multi-class classification, forcing the model to predict the entity identifier for each entity description in a training pair in addition to the match/non-match decision. Our evaluation across five entity matching benchmark datasets shows that dual-objective training can increase the matching performance for seen products by 1% to 5% F1 compared to single-objective Transformer-based methods, given that enough training data is available for both objectives. In order to gain a deeper understanding of the strengths and weaknesses of the proposed method, we compare JointBERT to several other BERT-based matching methods as well as baseline systems along a set of specific matching challenges. This evaluation shows that JointBERT, given enough training data for both objectives, outperforms the other methods on tasks involving seen products, while it underperforms for unseen products. Using a combination of LIME explanations and domain-specific word classes, we analyze the matching decisions of the different deep learning models and conclude that BERT-based models are better at focusing on relevant word classes compared to RNN-based models.

How reinforcement shapes the binding of stimulus-control associations

Cognitive control describes the ability to use internal goals to strategically guide how we process and respond to our environment. Changes in the environment lead to adaptation in control strategies. This type of control-learning can be observed in performance adjustments in response to varying proportions of easy to hard trials over blocks of trials on classic cognitive control tasks. Known as the list-wide proportion congruent (LWPC) effect, here, increased difficulty is met with enhanced attentional control. Recent research has shown that reinforcement events, in the form of performance feedback, enhance the LWPC effect, but the underlying mechanisms are not yet understood. To assess different hypotheses of how feedback is processed in the LWPC, we manipulated proportion congruency in a Stroop task over blocks of trials and provided trial-by-trial task-relevant word and task-irrelevant, trial-unique image performance feedback. The LWPC task was followed by a surprise recognition memory task for feedback images, which allowed us to test whether attention to feedback (incidental memory for the images) varies as a function of proportion congruency, time, and individual differences in reward sensitivity. We replicated a robust LWPC effect. Importantly, the memory data revealed better encoding of feedback images from context-defining trials (e.g., congruent trials in a mostly congruent block), especially early on in a new context, and in congruent conditions. Individual differences in reward sensitivity were not strongly associated with control-learning effects. These results suggest that reinforcement promotes the rapid forming of associations between stimuli and control demands, or context binding.

Do You Feel Me?

In working towards accomplishing a human-level acquisition and understanding of language, a robot must meet two requirements: the ability to learn words from interactions with its physical environment, and the ability to learn language from people in settings for language use, such as spoken dialogue. The second requirement poses a problem: If a robot is capable of asking a human teacher well-formed questions, it will lead the teacher to provide responses that are too advanced for a robot, which requires simple inputs and feedback to build word-level comprehension. In a live interactive study, we tested the hypothesis that emotional displays are a viable solution to this problem of how to communicate without relying on language the robot doesn't--indeed, cannot--actually know. Emotional displays can relate the robot's state of understanding to its human teacher, and are developmentally appropriate for the most common language acquisition setting: an adult interacting with a child. For our study, we programmed a robot to independently explore the world and elicit relevant word references and feedback from the participants who are confronted with two robot settings: a setting in which the robot displays emotions, and a second setting where the robot focuses on the task without displaying emotions, which also tests if emotional displays lead a participant to make incorrect assumptions regarding the robot's understanding. Analyzing the results from the surveys and the Grounded Semantics classifiers, we discovered that the use of emotional displays increases the number of inputs provided to the robot, an effect that's modulated by the ratio of positive to negative emotions that were displayed.

The BasisSpellingBank

The BasisSpellingBank is the first lexicon where the spellings and pronunciations of words are documented explicitly and separately for all relevant word parts. Unlike earlier descriptions of Dutch orthography in terms of rules and underlying forms, the BasisSpellingBank departs from the concept of storage and the way spelling is taught in schools. At its core are triplets of phoneme(s), grapheme(s), and the spelling category(s) which describe the correspondences between them. The triplet notation provides a detailed, exhaustive description of Dutch orthography. It is a formal system that could be used to describe other alphabetic writing systems as well. By integrating information about orthographic rules and lexical storage, the triplet notation more adequately describes the knowledge possessed by fluent users. The triplets unlock exact measures of both forward and backward consistency, which opens up detailed analyses of spelling performance. The database provides new insights into spelling education and spelling complexity.