Search

The article is an essay on the development of technologies for natural language processing, which formed the basis of BERT (Bidirectional Encoder Representations from Transformers), a language model from Google, showing high results on the whole class of problems associated with the understanding of natural language. Two key ideas implemented in BERT are knowledge transfer and attention mechanism. The model is designed to solve two problems on a large unlabeled data set and can reuse the identified language patterns for effective learning for a specific text processing problem. Architecture Transformer is based on the attention mechanism, i.e. it involves evaluation of relationships between input data tokens. In addition, the article notes strengths and weaknesses of BERT and the directions for further model improvement.

Статья посвящена коммуникативному методу обучения иностранным языкам, который носит деятельностный характер. Задача коммуникативного подхода заинтересовать обучаемых в изучении иностранного языка путём накопления и расширения их знаний и опыта. Основная цель этой методики научить школьника сначала свободно говорить на языке, а потом думать на нем., The article is devoted to the communicative method of teaching the foreign languages, which is the nature of the activity. The objective of the communicative approach is to interest students in learning a foreign language by the accumulation and expansion of their knowledge and experience. The main purpose of this procedure is to teach students to be fluent in the first language, and then to think on it. This is achieved due to the development of different types of monologues, dialogues and typical forms of language modeling.

Статья посвящена вопросам автоматического распознавания слитной речи. Проведен обзор современного состояния технологий распознавания, их применение в области транслирования и перевода речи в режиме реального времени. Рассматривается структура системы распознавания речи, включающая предварительную обработку речевого сигнала, акустическое моделирование, языковое моделирование и комбинирование. Особое внимание уделено этапу предварительной обработки сигналов, включающего выделение признаков речевого сигнала и их преобразование, в том числе выделение шумоустойчивых, адаптивных и дискриминативных характеристик. Приведены принципы построения таких акустических моделей, для которых применяются скрытые модели Маркова, оценка максимального правдоподобия, дискриминативное обучение.This paper is about the problems of continuous speech automatic recognition. We conducted a review of the current state of recognition technologies, their application in area of speech broadcasting and translating in real time. We considered the structure of speech recognition system, including front-end processing, acoustic modeling, language modeling and system combination. Special attention is paid to front-end processing, including speech feature extraction and transformation, in particular noise robust, speaker-adaptive and discriminative features. We presented principles of acoustic modeling building, for which applied hidden Markov models, maximum likelihood estimation, discriminative training, speaker adaptation, noise adaptation, deep neural networks.

В статье показана необоснованность разделения современных систем машинного перевода на SMT (Statistical Machine Translation), RBMT (Rule-Based Machine Translation), представлены доказательства, что только гибридные системы машинного перевода способны обеспечить адекватный перевод фраз и предложений. В статье изучены и проанализированы возможности машинного перевода, исследованы такие подходы, как SMT (Statistical Machine Translation), RBMT (Rule-Based Machine Translation), PBMT (Phrase-Based Machine Translation), EBMT (Example-Based Machine Translation), рассмотрены модели машинного перевода, их использование, с описанием принципа, функциональности, лингвистических характеристик, а также преимуществ и недостатков их использования. Для анализа использовались инструменты Mosestoolkit, TextStat, AntConc, современный корпус американского английского СОСА, русский национальный корпус и другие средства обработки текста., Considering machine translation, many authors treat separately Statistical Machine Translation and Rule-Based Machine Translation. The aim of the article is to prove the baselessness of such approaches. Machine Translation systems based on either of them can hardly cope with the translation of long, specific and complex strings. We consider that using the hybrid methods can lead to an adequate translation of sentences and phrases. The article aims to prove the inconsistence of treating Statistical Machine Translation and Rule-Based Machine Translation separately in machine translation systems, to analyze and study the possibilities of machine translation, different methods of text processing, models of machine translation, such as SMT (Statistical Machine Translation), RBMT (Rule-Based Machine Translation), PBMT (Phrase-Based Machine Translation), EBMT (Example-Based Machine Translation). Also the description their principle, functionality, linguistic performance, including their advantages and disadvantages are investigated. The following toolkits were used for the analysis: Moses toolkit, TextStat, AntConc, Contemporary American English Corpus СОСА, Russian National Corpus, The SRI language modeling toolkit and other tools for text processing. The article contains the principles of word normalization, alignment and sentence segmentation.

Рассмотрен вопрос о перспективах использования плановых языков в различных целях в качестве языка-посредника между различными народами и, в первую очередь, в качестве языка моделирования естественных языков. Описывается эксперимент по использованию конлангов как языка нетерминальных символов порождающих грамматик Хомского. Приведен анализ плановых языков на основе интерлингвистического подхода, связанного с изучением международных искусственных языков как средства преодоления языкового барьера. Делается вывод о том, что плановые языки межнационального общения могут найти свое применение не только в искусстве в качестве хобби, средства моделирования и анализа естественных языков, но и, в частности, для генерации осмысленной речи, осуществления языковых трансформаций и как языка нетерминальных символов порождающих грамматик., In the article the task is set to consider the issue of the prospects of using planned languages for different purposes as interlinguas between different nationalities and, first of all, as the natural language modeling languages. An experiment about the usage of conlangs as languages of non-terminal symbols of Chomsky’s generative grammars is described. The analysis is offered on the topic of planned languages based on interlinguistic approach, referred to studying international artificial languages as means of overcoming language barrier. A conclusion is made that the international communication with the help of planned languages can be applied not only as an art form, hobby and tools for modeling and analy-ing natural languages, but also particularly for natural language generation, executing linguistic transformations and as a generative grammar non-terminal symbols language.

The article provides teaching instructions for introductory session on poetic metaphor for EFL classrooms (pre-intermediate and above). Semiotic approach is used to classify metaphors. Special attention is given to encouragement of creative thinking and language modeling

В статье обосновывается актуальность создания пропозиционально-фреймового словаря телеутского языка. Моделирование структуры фрейма «Жилище» призвано продемонстрировать методологические принципы создания словарных статей. Описание фрейма «Жилище» отражает фрагмент языковой картины мира телеута представителя немногочисленного тюркоязычного народа, проживающего преимущественно в Кемеровской области., In the article, we prove the necessity to create the propositional framing dictionary of the Teleut language. Modeling of the structure of the frame intended to demonstrate the methodological principles of creating entries. A description frame “home” reflects the fragment of the Teleuts’ worldview who are members of a small Turkic people living mainly in Kemerovo region. Kemerovo region is home to about 2,500 Teleuts, of which 2,000 live in Gurievsky and Belovsky areas. A frame-based organization of the dictionary, where the operational units of the frame use a proposition, takes into account the associations between words, and the assignment to a specific fragment of extralinguistic reality. The tokens used in this article were collected as a result of expeditions in 2013-2014 in Bekovo village, Belovsky district of Kemerovo region. The source of tokens is used as a material of the Russian-Teleut dictionary. The article analyzes semantic nodes tokens verbalized frame “home”: the names of the home and its parts (the Teleut name of the house, the name of the house in size, the name of house number of hosts, the name of house number of floors, the names of the parts of the house, the names of the rooms in the house, the name of the home depending on what it is made), the name of outbuildings (the names of buildings for animals, the name of outbuildings by use), the name of building materials, he names of the owners of the home, the names of settlements. Frame structure is determined the nomadic life of the Teleuts for several centuries in the first place, and secondly, assimilation with Russian-speaking population. The Teleuts used in speech as traditional dwelling names and its parts are borrowed from the vocabulary of the Russian language, if they want to call the modern building materials, modern house parts, household items.

Статья посвящена проблеме моделирования речевых сбоев диктора при автоматическом распознавании речи. Рассматривается само явление речевых сбоев, и по результатам анализа отечественных и зарубежных работ выделяются две группы подходов к автоматическому определению речевых сбоев: параметрическая обработка сигнала и комбинированные методы с применением языкового моделирования. Приводится анализ особенностей их применимости к обработке русской речи., In this paper, the authors deal with the problem of speech disfluencies analysis for automatic speech recognition. The origin of speech disfluencies may be of different nature: it may be caused by external influence or by internal failure in the planning of speech act. Failures in the speech act planning may be various, among speech disfluencies one may name such as filled pauses, self-repairs and stipulations. Such disfluencies are an obstacle for automatic processing of speech and its transcriptions. Speech corpora with Rich Transcription (the transcription where such phenomena as sentence boundaries, fillers, disfluencies are marked) are used for studying speech disfluencies. Among such corpora are Czech Broadcast Conversation MDE Transcripts and SWITCHBOARD. It is still unclear what knowledge should be used in speech recognition systems to classify and elicit speech disfluencies. That is why there are no appropriate models of them, which could provide automatic disfluencies processing. Methods for such processing may be distinguished between those dealing with disfluencies by means of acoustic models and by means of combined models (acoustic and language ones). But owing to objective reasons (time and expert expenses) researchers frequently use only acoustic modeling in speech recognition systems. There are a lot of papers describing modeling of speech disfluencies as a part of ASR systems. There is also a group of approaches that are meant for speech recognition accuracy increase by separating disfluencies from speech signal in advance or by means of speech transcriptions. Among possible approaches to deal with these phenomena in ASR systems there are those that allow modeling and eliciting disfluencies as separate verbal and paralinguistic elements, and those that ignore them only distinguishing from useful speech, but not telling one from another. There is an alternative method of processing disfluencies as part of language modeling and modeling of unknown words: speech disfluencies may be treated as Unknown Words class, and then building a language model with the account of these phenomena. For the Russian language there are no methods developed for speech disfluencies processing, so it is worth trying to apply different methods and compare results. Due to high expenses of making a corpus of transcripts, which would account for speech disfluencies and would be suitable for training language model (at least 3-gram model), speech disfluencies processing with parametric methods seems to be optimal.

В статье рассматривается взаимодействие системы языка, ее моделирование через призму языковой личности, имеющей трехуровневую структуру, которая обусловливает моделирование языка и формирует языковую картину мира., The article studies the interaction of the system of language through its modeling by means of the three level phenomenon of language personality, which determines language modeling and thus forms the world view.

В статье излагаются основные методологические взгляды автора на возможность сублогического моделирования языка в рамках актуальной лингвистической парадигмы XXI века когнитивной лингвистики. Опираясь на междисциплинарность когнитивного подхода к языку, привлекая новейшие разработки из области философии языка, автор обращается, во-первых, к характеристике таких процессов, как относительная трансцендентность, реактивность, выход в измененные состояния сознания, деконструкция, тропность, а во-вторых, к вопросу о возможности моделирования феномена языка на сублогическом уровне., The article touches upon the author's basic methodological views of subconscious language modeling within cognitive linguistics. Given the ambiguity of cognitive approach to language and latest achievements in the philosophy of language such concepts are tackled as transcendency, conditioned nature, metaphoric character, altered states of consciousness, deconstruction and the possibility of modeling the phenomenon at subconscious level.