Attention graph: Learning effective visual features for large-scale image classification

In recent years, the research of deep learning has received extensive attention, and many breakthroughs have been made in various fields. On this basis, a neural network with the attention mechanism has become a research hotspot. In this paper, we try to solve the image classification task by implementing channel and spatial attention mechanism which improve the expression ability of neural network model. Different from previous studies, we propose an attention module consisting of channel attention module (CAM) and spatial attention module (SAM). The proposed module derives attention graphs from channel dimension and spatial dimension respectively, then the input features are selectively learned according to the importance of the features. Besides, this module is lightweight and can be easily integrated into image classification algorithms. In the experiment, we combine the deep residual network model with the attention module and the experimental results show that the proposed method brings higher image classification accuracy. The channel attention module adds weight to the signals on different convolution channels to represent the correlation. For different channels, the higher the weight, the higher the correlation which required more attention. The main function of spatial attention is to capture the most informative part in the local feature graph, which is a supplement to channel attention. We evaluate our proposed module based on the ImageNet-1K and Cifar-100 respectively. Through a large number of comparative experiments, our proposed model achieved outstanding performance.

Kommunikative und expressive Formeln des Deutschen in Internettexten: ein diskursorientierter Ansatz

This paper deals with the functioning of expressive and communicative formulas of modern German in computer-mediated discourse. The set phrases analysed are pre-formed sentence-value word combinations that serve as a means of expression of various speech intentions and emotions. These linguistic units are known to evade a uniform classification scheme, with boundaries between different groups being indistinct. The set phrases considered in this work are viewed as a functional-semantic field which includes communicative and expressive formulas and some mixed types. By discussing their cultural specificity – provided by units containing symbolic components, as well as by entities in various areas of national life, and fixed phrases of obscure literary origin – I attempt to explore their intertextual potential for further contribution to studies in intercultural communication. Expressivity, intertextuality, multimedia impact and interactivity have been defined as constitutive characteristics of computer-mediated discourse. The units discussed keep the tone of asynchronous online communication humorous and help to establish communicative closeness. In online media, the language game comes to the fore: a common phenomenon is semantic transformation often performed as simultaneous actualisation (ambiguity) of literal and idiomatic meaning, causing additional communicative effects. In combination with the informative part of a headline, these units act as emotional and evaluative indicators of media texts. Due to its technological features, internet communication allows easy combination of visual and verbal channels of information, and so there is frequent play between language images and digital images, especially in photo and video hosting services. My analysis of different online genres has shown that in internet communication, the core of the functional-semantic field (i.e. the most frequently used units) acquires added discursive value. The universal nature of their creative use, appeal, entertainment and community-building functions was also ascertained.

Answer Selection in Community Question Answering Using LSTM

Websites like Quora, Yahoo! Answers, and Reddit are examples of community question answering (CQA) systems that enable users to ask questions as well as to answer questions. Answer selection is the most challenging task in CQA systems to get the good and relevant answer for the user questions. The shortcomings in the current approaches are lexical gap between text pairs, dependency on external sources, and manual features which lead to lack of generalization ability. These shortcomings are resolved by already proposed work, but they lack generalization, and their performance is not satisfying. Whereas to focus on rich quality answers, attention mechanism can be integrates with neural network. This chapter proposes two models BLSTM and BLSTM with attention mechanism. Attention mechanism aligns question to the answer with the answer's more informative part. So, when it is applied in the model, BLSTM with attention mechanism model surpasses the top approaches.

FEATURES OF TRANSLATION OF COMPLEX SENTENCES WITH SEMANTICS OF INDIRECT EVIDENTIALITY FROM FRENCH INTO RUSSIAN

The article is devoted to the analysis of complex sentences with indirect evidentiality semantics in French. A complex sentence is an explicit way of expressing indirect evidentiality. Indirect evidentiality implies the receipt of information from third parties. A complex sentence is a mean of complete representation of information source. As a result of the study it was found that in French indirect evidentiality is expressed by complex sentences. The main clause indicates information source. The dependent clause introduced by conjunction que is informative. The subordinate clause in majority of cases is in postposition of the main clause. French complex sentence is translated into Russian by analogue complex sentence. Only in some cases it is possible to replace the subordinate clause as a verbally expressed informative part by a process noun. In French and in Russian informing verbs are involved at the lexical level. In French the verbs informer, announcer are involved. Occasionally the verb renseigner is used. In translation these verbs correspond to Russian verbs of similar semantics: декларировать / to declare, сообщать / to report, извещать / to inform.

EPITAPHIES AS A HISTORICAL AND ETHNOGRAPHIC SOURCE (ON THE EXAMPLE OF VILLAGES. DAGBASH/AKHWAKH, SHAMIL DISTRICT, REPUBLIC OF DAGHESTAN)

Based on analysis of the information contained in the newly discovered Arabic language epitaphs, the article shows the efficiency of the integrated approach in historical and ethnographic researches. This study was carried out by the authors on the example of the village of Dagbash (Akhwakh), the Shamil district of the Republic of Daghestan. The epitaphs of the late XV – early XIX centuries have been studied. They were found on the grave steles of the settlement cemetery and on a stone inserted in the wall of the mosque. The informative part of the epitaphs is analyzed together with the oral traditions, ethnographic observations, as well as data from the Arabic-language written sources. Such an integrated approach made it possible for the authors to date the formation of this rural community, as well as to trace the geography of the migration processes that accompanied this phenomenon. The authors conclude that this community formed before the XV century, as a result of the unification of three more similar settlements located nearby the existing Akhwakh settlement. Later, several more clans (tukhums) formed in Akhwakh, which arose during the sprawl of families of immigrants from other settlements. The clan structure of Akhwakh apparently had formed by the beginning of the XIX century. This approach is quite applicable to studies of similar examples of synoecism in medieval Daghestan. Also, the analysis of epitaphs made it possible to trace the evolution of anthroponomy in the given microregion, as well as its dependence on a change in the religious affiliation of its population. The authors also managed to identify in written sources the original name of the village, which was changed in the 1932 (Akhwakh to Dagbash).

An Efficient Framework for Dense Video Captioning

Dense video captioning is an extremely challenging task since an accurate and faithful description of events in a video requires a holistic knowledge of the video contents as well as contextual reasoning of individual events. Most existing approaches handle this problem by first proposing event boundaries from a video and then captioning on a subset of the proposals. Generation of dense temporal annotations and corresponding captions from long videos can be dramatically source consuming. In this paper, we focus on the task of generating a dense description of temporally untrimmed videos and aim to significantly reduce the computational cost by processing fewer frames while maintaining accuracy. Existing video captioning methods sample frames with a predefined frequency over the entire video or use all the frames. Instead, we propose a deep reinforcement-based approach which enables an agent to describe multiple events in a video by watching a portion of the frames. The agent needs to watch more frames when it is processing an informative part of the video, and skip frames when there is redundancy. The agent is trained using actor-critic algorithm, where the actor determines the frames to be watched from a video and the critic assesses the optimality of the decisions taken by the actor. Such an efficient frame selection simplifies the event proposal task considerably. This has the added effect of reducing the occurrence of unwanted proposals. The encoded state representation of the frame selection agent is further utilized for guiding event proposal and caption generation tasks. We also leverage the idea of knowledge distillation to improve the accuracy. We conduct extensive evaluations on ActivityNet captions dataset to validate our method.

Headings in Thomas Brussig’s novel «Am kürzeren Ende der Sonnenallee» and their reproduction in the ukrainian translation

The paper explores the peculiarities of headings in Thomas Brussig’s novel «Am kürzeren Ende der Sonnenallee» and the means of reproducing their functions in the Ukrainian translation. The problems of equivalency and subjective transformations in connection with author’s style are investigated. Headings as strong positions are important elements of texts architectonic and specific communicative unities. We distinguish many functions such as nominative, informative, expressive, emotive functions which dominate according to the genre and style of texts. Scientists pay much attention to such aspects as semantics, classification, structure and functions of headings. Different types of headings need different ways of their translation. Equivalents in the functions are more important than formal equivalency. The informative expository headings are easy to reproduce in translation by using the same descriptive models. In the novel there are only a few topic-identifying headings where the content is explained, so that the translation has the functional equivalency with the original. It is more difficult when headings have intertextual character and contain many allusions instead of direct content description. Brussig’s chapter headings only partially reflect the text content and let the reader interpret it in different way. It depends on general cultural context (cultural lacunes) and background knowledge if all the intertextual elements can be understood. That’s why it often seems necessary to choose the explanatory headings while translating. They make headings more transparent but there is no ambiguity and disappointed expectation due to the original. In the article has been shown how the associations connected with song names, literary quotes, polysemantic words attract the attention of readers and influence the interpretation of chapter content and which effects disappear by translation. In the Ukrainian translation all the headings have an informative part containing the words «a story about» as a marker for content. The choice of topic-identifying headings for translation causes great differences in the interpretation of chapters because of predictability missing in the original. It is especially undesirable when other associations appear by reading of the translation. The variation in the length of the text section corresponds to a difference in the scope of the information in the translation. In long chapters there are many acts and events but the translator chooses subjectively only one of them for heading content. In this case the heading has a strong steering role for the reader and the functional equivalency is lost.

THE USE OF DOCUMENT AND INFORMATION TOOLS IN THE FORMATION OF A MODERN MANAGEMENT SYSTEM OF THE ORGANIZATION

The article is devoted to research basic aspects of formation, development and implementation of the management Organization of documents and information facilities, forming with their help a comprehensive system of documentary support of Enterprises and Organizations. Implementation of management functions is directly related to information processes. The document is one of the forms of information participation in the communicative process. Information is acquires the necessary qualities and properties: authenticity, evidence, effectiveness and can be stored and resubmitted in space and time in a documented form. Research in the field of records keeping allows to consider the document as a complex information system, which is characterized by many parameters and is in constant development. The method of typing (or stencil) texts is one of the most studied methods unification the content of documents. It received the largest distribution in the unification of organizational and administrative documentation, where the informative part of documents is laid out in the text form. The historical development of the existing system of documentation of management information and documenting the activities of organizations, institutions and enterprises has resulted in creating the whole system of documentation and management of documents, which was formed in Separate activity. Building an effective management system for enterprises, organizations, processes and types of activity is most the value of compliance with certain rules in documenting the activities of enterprises and organizations is enhanced in the modern world and is caused by an increase in attention on it. "The document management" have includes the following functions and tasks: adoption of policies and standards in the field of document management; Allocation of responsibilities and powers to manage documents; Establishing and distributing procedures and guidelines for managing documents in the organization; Providing a range of services that relate to the management of documents and the use of documents; Design, implementation and administration of specialized document management systems; Integration management of documents in the system and processes of business activity.

The role of the notes to the financial statements in corporate decision-making

For the analyses, we used the notes to the financial statements of companies with Hungarian headquarters, which are regulated by the Accounting Act and which have information-technology services as their primary activity of the business. The primary objective of our research was to examine the extent to which the enterprises investigated fulfill their obligations determined by the Accounting Act to disclose all needed information in the general and informative parts of the notes to the financial statement. The analysis was performed using text mining, and the results were evaluated with non-parametric statistical methods. Based on the analysis, we can state that the investigated companies presented proportionally less information in the general part of the notes to the financial statements as compared to the informative part. However, even the highest frequency did not exceed 70%. Based on the χ2 test, we can state that the frequency of occurrence of the majority of information sought using text mining significantly differs by company type, and this difference mostly occurred amongst Ltds and Lps, in favor of Ltds. As a conclusion, it can be deduced that the investigated companies do fully not meet the legal requirements for the information provided in the notes to the financial statements. This incompleteness may cut down on the decision-making ability of the companies concerned.