This chapter surveys the history and main directions of natural language processing research in general, and for Slavic languages in particular. The field has grown enormously since its beginning. Especially since 2010, the amount of digital texts has been rapidly growing; furthermore, research has yielded an ever-greater number of highly usable applications. This is reflected in the increasing number and attendance of NLP conferences and workshops. Slavic countries are no exception; several have been organising international conferences for decades, and their proceedings are the best place to find publications on Slavic NLP research. The general trend of the evolution of NLP is difficult to predict. It is certain that deep learning, including various new types (e.g. contextual, multilingual) of word embeddings and similar ‘deep’ models will play an increasing role, while predictions also mention the increasing importance of the Universal Dependencies framework and treebanks and research into the theory, not only the practice, of deep learning, coupled with attempts at achieving better explainability of the resulting models.

The introduction to this volume describes its content. It also provides the rationale for including selected topics and provides comments on the manner of presentation adopted in this volume.

In this chapter, a case is made for the inclusion of computational approaches to linguistics within the theoretical fold. Computational models aimed at application are a special case of predictive models. The status quo in the philosophy of linguistics is that explanation is scientifically prior to prediction. This is a mistake. Once corrected, the theoretical place of prediction is restored and, with it, computational models of language. The chapter first describes the history behind the emergence of explanation over prediction views in the general philosophy of science. It’s then suggested that this post-positivist intellectual milieu influenced the rejection of computational linguistics in the philosophy of theoretical linguistics. A case study of the predictive power already embedded in contemporary linguistic theory is presented through some work on negative polarity items. The discussion moves to the competence–performance divide informed by the so-called Galilean style in linguistics that retains the explanatory over prediction ideal. In the final sections of the chapter, continuous methods, such as probabilistic linguistics, are used to showcase the explanatory and predictive possibilities of nondiscrete approaches, before a discussion of the contemporary field of deep learning in natural language processing (NLP), where these predictive possibilities are further amplified.

This chapter investigates why Pāṇini has been called the world’s first computational linguist and situates Pāṇinian grammar within the context of formal language theory and issues of generative capacity, which have been of huge importance in the development of modern linguistic theory since Chomsky's earliest work. Existing claims regarding the generative capacity of Pāṇini’s system are shown to be inadequate, and a new assessment of the power of the Aṣṭādhyāyī is provided.

Why do successive education reforms within a country resonate with familiar assumptions about educational goals, society, class, and state, even at moments of radical change? Repeating cultural narratives sustain continuities within institutional change processes, by influencing how new ideas are interpreted, how interest groups express preferences, and how institutional norms shape political processes. Repeating narratives make it more likely for some types of reforms to be implemented and sustained than others. This chapter develops a theoretical model suggesting how cultural narratives are transmitted across time and an empirical method for assessing cross-national differences in cultural narratives. Each country has a distinctive “cultural constraint,” or a set of cultural symbols and narratives, that appears in a nation’s literary corpus. Writers collectively contribute to this body of cultural tropes; despite individual fluctuations, they largely reproduce the master narratives of their countries. Computational linguistic processes allow us to observe empirical differences between British and Danish cultural depictions of education in 1,084 works of fiction from 1700 to 1920. Cultural narratives do not determine specific outcomes, as tropes must be activated in political struggles. Yet we can show how significant cross-national differences in literary images of education resonate with British and Danish educational trajectories.

This chapter introduces our global dataset of autocratic propaganda, which contains over eight million articles from 65 newspapers drawn from 59 countries in six major languages. By population, our dataset encompasses a set of countries that represents 88\% of all people who live under autocracy. After collecting this propaganda, we measured its content. We employ computational techniques to identify the topics of each article; count the number of references in each article to the autocrat, ruling party, and opposition; and measure the valence of propaganda with dictionary based semantic analysis. The key idea is that some words have an intrinsic positive or negative sentiment. This conception of propaganda -- as spin, not lies -- accords with how scholars and practitioners have long understood it. As a baseline for comparison, our dataset includes state-affiliated newspapers from democracies. To scale our measures of propaganda, we develop a Fox News Index: how Fox News covers Republicans relative to Democrats.

The distinction that has become standard between natural language and formal language, which rests on differentiating what is socially evolved and experiential from what is purposefully planned, suggests that a similar emphasis on experientiality may illuminate the distinction between narrative and formal modes of knowing, which figures prominently in this volume. Support for that perspective comes from developments in both narratology and computational linguistics. A key concept from both specialties – and for this volume – is that of ‘scripts’, which indicates how even texts that are explicitly formal may be understood as narratives by experienced readers. An explicit example that illuminates these themes comes from James Clerk Maxwell’s classic paper ‘On Faraday’s Lines of Force’. It juxtaposes narrative and formal modes of representation and displays their relative advantages, suggesting that the development of scientific knowledge often depends on continual feedback between natural narrative and formal analysis.

This chapter presents an easily followed overview of computational linguistics and where Arabic fits into it. Computational linguistics, often referred to interchangeably as natural language processing (NLP) or human language technologies, is a large and growing interdisciplinary field of research that lies at the intersection of linguistics, computer science, electrical engineering, cognitive science, psychology, pedagogy, and mathematics, among other fields. Research and work on Arabic computational linguistics has lagged behind English and other languages. This is despite a tremendous increase in the relative growth of Arabic NLP in the period between 2012 and 2016. The reason for its slow start is that Arabic presents a series of difficulties to programmers, those being morphological richness, orthographic ambiguity, dialectal variations, orthographic noise, and resource poverty. Those problems have been or are being overcome, and a new generation of researchers has made great strides in the field. This has partly to do with the growing interest in language technologies for opinion mining and translation in social media, which features dialectal Arabic more than MSA. Another motivation is that commercial giants like Apple and Google are interested in applications of Arabic as it is spoken.

The authors examine the application of electronically searchable corpora, from their own experience, in addressing questions pertinent to linguistics as a whole and to matters internal to Arabic, the while lamenting that the field of Arabic linguistics, in its theoretical and applied orientations alike, has not made use of the rich data source that searchable electronic corpora represent. They show how corpora can be used easily to falsify common assumptions and assertions about the human language capacity in general just as they can be used efficiently to query assumptions and assertions about Arabic itself. So, too, do they hold implications for applied uses such as teaching Arabic as a foreign language and translation between Arabic and other languages. In any of these applications, the use of corpora in the analysis of all varieties of Arabic remains underdeveloped compared to their use in the analysis of other languages, especially English.

This chapter illustrates how computational models of ISA disambiguation havebeen improved and eventually applied to human-like robots in dailyinteractions, taking into account contextual parameters such as uncertaintyabout a speaker's intentions, hierarchical status and politenessexpectations.