Features Of Natural Language Research Articles

The language of music: Common neural codes for structured sequences in music and natural language

The ability to process structured sequences is a central feature of natural language but also characterizes many other domains of human cognition. In this fMRI study, we measured brain metabolic response in musicians as they generated structured and non-structured sequences in language and music. We employed a univariate and multivariate cross-classification approach to provide evidence that a common neural code underlies the production of structured sequences across the two domains. Crucially, the common substrate includes Broca’s area, a region well known for processing structured sequences in language. These findings have several implications. First, they directly support the hypothesis that language and music share syntactic integration mechanisms. Second, they show that Broca’s area is capable of operating supramodally across these two domains. Finally, these results dismiss the recent hypothesis that domain general processes of neighboring neural substrates explain the previously observed “overlap” between neuroimaging activations across the two domains.

Semi-automatic sentiment analysis based on topic modeling

Sentiment is an important feature of natural language. It is used to understand semantic of texts and opinion of people. There are many practical applications, which require to extract sentiment from texts: advertising analytics, interactive chat bots, opinion mining. Today, different supervised techniques are used to extract sentiment from texts which require large manually labeled datasets that are expensive and time consuming to build. Moreover, such datasets should cover vocabularies and patterns of use of different contexts. Additionally, the efficiency of supervised methods trained on a well-written texts can dramatically decrease on users’ texts from social media due to typos, slang, short length of sentences.To solve these problems and to reduce human involvement, we propose semi-supervised sentiment analysis method based on topic modeling with Additive Regularization. To evaluate the efficiency of this method we applied it to several open-source datasets for which sentiment labels are available. The study shows promising results in terms of f1-score with minimal human involvement.

The Design and Implementation of a Covering MDN-Complete-Life-Cycle Malicious Domain Detection Framework

Malicious domain name (MDN) detection has seen greatly progress in recent years. In this paper, one covering MDN-Complete-Life-Cycle malicious domain name detection framework is proposed. The framework includes three detection models: DGAD-M (Domain Generation Algorithm Detection Model), DIPD-M (Domain IP Detection Model) and DHTD-M (Domain Host Detection Model), corresponding to the process of the malicious domain generation, malicious domain name resolution and the host requesting a domain. DGAD-M bases on the fact that the domains generated by DGA are always short of natural language features, it adopts Convolutional Neural Network. DIPD-M bases on the fact that the IP addresses of the malicious domains are more disperse and updated frequently. DHTD-M bases on the fact that the domains requested by infected hosts are frequently tend to be malicious. The results of DGAD-M and DIPD-M will be used by DHTD-M. The framework got the accuracy rate of 83.652% with the real network flow and found out 115 suspicious malicious domains.

Of Gaps and Holes and Silence: Some Remarks on Elliptic Speech and Pseudo-Orality in James Joyce’s Short Story “The Sisters”

This paper discusses aspects of direct speech in James Joyce’s story “The Sisters”. The story is often analyzed with special attention to the gaps and ellipses in the utterances, which are usually read as omissions, evasions, or uncomfortable silences, and thus as indicative of some transgressive behaviour of the dead priest who is at the centre of the dialogues. In this article we explore the hypothesis that the utterances in question show features that are quite common in natural spoken language and thus may also be read as literary techniques to create authentic oral discourse. This hypothesis is not intended to invalidate previous interpretations, but to introduce an additional aspect of interpretation that has been neglected so far. In the context of a literary work, features of natural spoken language acquire new meaning, and the very attempt to narrow the gap between literary and natural spoken language appears as inauthentic, ominous and as an artistic strategy to express the unspeakable. The story thus evokes suspicion and a feeling of eeriness while also offering narrative and linguistic information that allows for a more empathic assessment of the characters. We use quantitative methods of analysis and linguistic data from corpora of (authentic) spoken language to substantiate our hypothesis. As “The Sisters” is a rather short story and the present study is, in several respects, exploratory, our claims and hypotheses need to be confirmed and validated by more exhaustive research into Joyce’s major works.

Vagueness and Aggregation in Multiple Sender Channels

Vagueness is an extremely common feature of natural language, but does it actually play a positive, efficiency enhancing, role in communication? Adopting a probabilistic interpretation of vague terms, we propose that vagueness might act as a source of randomness when deciding what to assert. In this context we investigate the efficacy of multiple sender channels in which senders choose assertions stochastically according to vague definitions of the relevant words, and a receiver then aggregates the different signals. These vague channels are then compared with Boolean channels in which assertions are selected deterministically based on classical (crisp) definitions. We show that given a sufficient number of senders, a linear stochastic channel outperforms Boolean channels when performance is measured by the expected squared error between the actual value described by the senders and the receiver’s estimate of it based on the signals they receive. The number of senders required for vague channels to be at least as accurate as Boolean channels is shown to be a decreasing function of the size of the language i.e. the number of description labels available to the senders. Vague channels are then shown to be robust to transmission error provided the error rate is not too large. In addition, we investigate the behaviour of both Boolean and vague channels for a parametrised family of distributions on the input values. Finally, we consider optimal vague channels assuming a fixed number of senders and show that, provided there are more than two senders, a vague channel can be found that outperforms the optimal Boolean channel. In this context, we show that for channels with relatively low numbers of senders S-curve production functions are optimal.

Hybrid Interpreting Framework of the Unified Scheduling Language

The unified scheduling language is a Chinese domain-specific language with both programming and natural language features for space mission scheduling. To create a functional and yet simple language for all space-mission-related staff, a hybrid framework for language defining and interpreting is proposed. This framework combines script programming languages and natural language processing technologies, and it creates a language that is easy to learn yet still powerful to fulfill the automating and extending requirements of mission scheduling. A coordinated natural language processing approach is designed to parse human-oriented languages, and a general-purpose script engine is integrated for processing the script aspect of the unified scheduling language. The translating mechanism maintains logic consistency between the two parsers. Related logic models and algorithms are introduced to illustrate the parsing mechanism. This framework achieves sound practical results in the unified scheduling language, considering a valid domain-specific language system is built and parsed correctly. The framework can be generalized to other Chinese domain-specific languages in various application fields.

Recurrent Neural Networkによるマイクロブログの投稿位置推定

People collect and use information about real world from internet to help their daily activities. In particular, the number of users in microblog such as Twitter is so large that users can get a diversity of information. They can elicit not only the information which they need from microblog posts but also the location which is indicated by the contents posted in microblog. While previous approaches apply corpus-based or machine learning that require various prior knowledge such as natural language processing and feature engineering, our approach is able to estimate the location without those requirements with extension of long-short term memory (LSTM). In our experiment, we apply our approach to geo-tagged tweets posted in Twitter and show that this approach is effective in outperforming corpus-based and previous works that use support vector machine (SVM) with bag-of-words (BoW).

The regularity game: Investigating linguistic rule dynamics in a population of interacting agents

Rules are an efficient feature of natural languages which allow speakers to use a finite set of instructions to generate a virtually infinite set of utterances. Yet, for many regular rules, there are irregular exceptions. There has been lively debate in cognitive science about how individual learners acquire rules and exceptions; for example, how they learn the past tense of preach is preached, but for teach it is taught. However, for most population or language-level models of language structure, particularly from the perspective of language evolution, the goal has generally been to examine how languages evolve stable structure, and neglects the fact that in many cases, languages exhibit exceptions to structural rules. We examine the dynamics of regularity and irregularity across a population of interacting agents to investigate how, for example, the irregular teach coexists beside the regular preach in a dynamic language system. Models show that in the absence of individual biases towards either regularity or irregularity, the outcome of a system is determined entirely by the initial condition. On the other hand, in the presence of individual biases, rule systems exhibit frequency dependent patterns in regularity reminiscent of patterns found in natural language. We implement individual biases towards regularity in two ways: through ‘child’ agents who have a preference to generalise using the regular form, and through a memory constraint wherein an agent can only remember an irregular form for a finite time period. We provide theoretical arguments for the prediction of a critical frequency below which irregularity cannot persist in terms of the duration of the finite time period which constrains agent memory. Further, within our framework we also find stable irregularity, arguably a feature of most natural languages not accounted for in many other cultural models of language structure.

On Something Similar

The use of expressions something similar/something like that/and the like is a pervasive feature of spoken natural language, and thus must be accounted for if we hope to develop a semantic account of dialogue. These expressions exhibit an anaphoric type dependency, but at the semantic-pragmatic level rather than the syntactic level. The meaning of these utterances depends upon the context of utterance: if the meaning of the utterance is regardedas fixed, then utterances which intuitively are coherent receive incoherent interpretations. To model what happens with utterances containing these phrases, we require both a formal theory of discourse structure, and accessibility to the lexicon.

Reconciling NSM and Formal Semantics

Formal semantics and Natural Semantic Metalanguage are widely held to be radically incompatible as ways to study meaning in naural language. Here I will show that they can to some extent be reconciled. In particular, for linguists working with NSM, formal semantics can be viewed as providing mathematical accounts of some of the same phenomena that NSM studies, such as entailment, and for the formal semanticist, NSM offers a small target in the form of mini-languages that exhibit the essential logico-semantic features of full natural languages, such as extensionality, intensionality and hyperintensionality, and algebraic principles such as transitivity, symmetry etc. or various of the primes. Therefore, although these two approaches are likely to remain distinct enterprises for the foreseeable future, some intercommunication is possible and indeed desirable.

Prior’s individuals

Criticisms have been aired before (see Bealer 1982, Schiffer 2003 for example) about the fear (which Arthur Prior shared) of certain Platonic abstract objects, propositions. That criticism extends to the widespread preference (which Prior initiated) for an operator analysis of expressions like ‘It is true, known, obligatory that p’ as opposed to the predicative analysis in their equivalents ‘That p is true, known, obligatory’. The criticism in the present work also concerns Prior’s attitude to Platonic entities of a certain kind: not propositions, i.e., the referents of ‘that’-clauses, but individuals, i.e., the referents of Russell’s ‘logically proper names’. Prior had a close knowledge of theories of individuals, including that of Russell. However, when Prior formulated his tense logic Q he presumed a non-Russellian, pre-suppositional account of individuals, such that they might exist at one time but not at another, and when they did not exist no statement could be made about them. This paper examines the correctness of such a pre-suppositional account of individuals, showing there are features of natural language it cannot account for. It is shown that the required features can be formalised in Hilbert’s epsilon calculus, and that using that calculus substantiates a Russellian understanding of ‘logically proper names’.

Riemannian metrics for neural networks II: recurrent networks and learning symbolic data sequences

Recurrent neural networks are powerful models for sequential data, able to represent complex dependencies in the sequence that simpler models such as hidden Markov models cannot handle. Yet they are notoriously hard to train. Here we introduce a training procedure using a gradient ascent in a Riemannian metric: this produces an algorithm independent from design choices such as the encoding of parameters and unit activities. This metric gradient ascent is designed to have an algorithmic cost close to backpropagation through time for sparsely connected networks. We use this procedure on \emph{gated leaky neural networks} (GLNNs), a variant of recurrent neural networks with an architecture inspired by finite automata and an evolution equation inspired by continuous-time networks. GLNNs trained with a Riemannian gradient are demonstrated to effectively capture a variety of structures in synthetic problems: basic block nesting as in context-free grammars (an important feature of natural languages, but difficult to learn), intersections of multiple independent Markov-type relations, or long-distance relationships such as the distant-XOR problem. This method does not require adjusting the network structure or initial parameters: the network used is a sparse random graph and the initialization is identical for all problems considered.

Vagueness and Imprecision: Empirical Foundations

Vagueness is a pervasive feature of natural language, which has been studied from a range of perspectives. This review focuses on recent empirical insights into vagueness that have come out of the field of linguistic semantics, as well as the theoretical developments that these have prompted. Topics covered include the distinction between vagueness and imprecision, or what I refer to as Type 1 and Type 2 vagueness; the complex manifestations of vagueness in the adjectival domain; and recent experimental findings regarding “ordinary” speakers’ use and interpretation of vague language. Also briefly discussed is the broader question of why language is vague.

Life is physics and chemistry and communication.

Manfred Eigen extended Erwin Schroedinger's concept of "life is physics and chemistry" through the introduction of information theory and cybernetic systems theory into "life is physics and chemistry and information." Based on this assumption, Eigen developed the concepts of quasispecies and hypercycles, which have been dominant in molecular biology and virology ever since. He insisted that the genetic code is not just used metaphorically: it represents a real natural language. However, the basics of scientific knowledge changed dramatically within the second half of the 20th century. Unfortunately, Eigen ignored the results of the philosophy of science discourse on essential features of natural languages and codes: a natural language or code emerges from populations of living agents that communicate. This contribution will look at some of the highlights of this historical development and the results relevant for biological theories about life.

On What Actually Is

The actually-operator, understood as a rigidifier, has been employed for a range of purposes in natural language semantics. In this article I argue that the properties of the operator do not correspond to any feature of natural language or feature natural language users have access to. Nor is it needed to provide a formal representation of natural language sentences—the examples usually provided to illustrate the indispensability of the operator are much more plausibly interpreted using plural quantifiers. This lack of connection to natural language is a serious worry for accounts that appeal to rigidifying operations to explain natural language phenomena, as well as a challenge to theories that appeal to the operator to capture the difference between different kinds of necessity expressed in natural language.

THE QUANTIFIED ARGUMENT CALCULUS

AbstractI develop a formal logic in which quantified arguments occur in argument positions of predicates. This logic also incorporates negative predication, anaphora and converse relation terms, namely, additional syntactic features of natural language. In these and additional respects, it represents the logic of natural language more adequately than does any version of Frege’s Predicate Calculus. I first introduce the system’s main ideas and familiarize it by means of translations of natural language sentences. I then develop a formal system built on these principles, the Quantified Argument Calculus or Quarc. I provide a truth-value assignment semantics and a proof system for the Quarc. I next demonstrate the system’s power by a variety of proofs; I prove its soundness; and I comment on its completeness. I then extend the system to modal logic, again providing a proof system and a truth-value assignment semantics. I proceed to show how the Quarc versions of the Barcan formulas, of their converses and of necessary existence come out straightforwardly invalid, which I argue is an advantage of the modal Quarc over modal Predicate Logic as a system intended to capture the logic of natural language.

Automatic evaluation scheme for imperceptibility of natural language watermarking

The deficiency of evaluation scheme for the imperceptibility of Natural Language Watermarking(NLW)seriously restricts the development in this field. To resolve this problem,combining the characteristics of natural language and the knowledge of psycholinguistics and considering the difficulty for human to understand natural language,an imperceptibility perceptual model and an automatic evaluation scheme were proposed. The proposed scheme evaluated the imperceptibility from three aspects: syntax error,collocation error and semantic loss. Finally,four different watermarking schemes including the TLex watermarking system based on absolute synonym substitution,the relative synonymous substitution watermarking system,the watermarking system based on syntactic tree and the Chinese syntactic watermarking system were used to compare the results of the proposed automatic method and manual method whose confidence coefficient is 90%. The two methods get the same conclusion: the imperceptibility of NLW based on word is better than that of NLW based on sentence,and the automatic evaluation scheme is effective for the imperceptibility of NLW.

Semantic Characterization of Tweets Using Topic Models

In the entertainment domain users tweet about their expectations and opinions regarding upcoming, current and past experiences, while companies advertise and promote the shows. This characterization, important for customers and companies, goes beyond traditional sentiment analysis where the polarity of the sentiments expressed in opinions is usually identified as positive, negative or neutral. The authors investigate different tweet representation models, including bags of words and probabilistic topic models, to shed light on the semantics of the messages. Their experiments show that topic-based models generated with Latent Dirichlet Allocation (LDA) yield, most of the times, better categorizations when compared to TF-IDF based features, particularly when these models are enriched with natural language features and specific Twitter slang.

The Significance of Behaviour-Related Criteria for Textual Exegesis—and Their Neglect in Indian Studies

Against the background of the fact that speakers not seldom intend to convey imports which deviate from the linguistically expressed meanings of linguistic items, the present article addresses some consequences of this phenomenon which appear to still be neglected in textual studies. It is suggested that understanding behaviour is in some respect a primary objective of exegesis and that due attention must be attributed to the high diversity of behaviour-related criteria by which interpretations of linguistic items are to be evaluated. Although we intimate in addition that individual (meaningful) sentences occurring either in oral conversations or in written documents generally exhibit a multiplicity of contents of diverse types and that the circumstance that sometimes only a content equalling the linguistic significance of a pertinent unit matters for purposes of interpretation is caused by a material coincidence of different varieties of content, the tenets advocated in the paper do not essentially depend on that view. On the other hand, the following assumptions are relevant in the present connection: (a) A number of deviances between imports conveyed by linguistic utterances and literal meanings of expressions occur due to maxims of linguistic behaviour that are quite independent of lexical and syntactic features of individual natural languages. (b) It is by no means an exceptional phenomenon that imports not derivable by grammatical rules of a particular language alone possess primary importance for interpretation and textual exegesis. In view of significant affinities between understanding of sentences and of texts it is argued that the consideration of diverse aspects of behaviour possesses relevance for textual exegesis at least in the following respects: (1) By delivering a heuristic device for discerning problems affecting adopted interpretations it encourages searches for alternatives. (2) It provides means for evaluating the degree of acceptability of particular textual exegeses and possibly rejecting them on a more rational basis than mere intuition. (3) It offers possibilities for critically assessing the validity of explicit arguments advanced in favour of or in opposition to some interpretation. (4) It furnishes a background for assessing certain disputes about translation. The dimension of linguistic behaviour also attains importance in connection with questions of exegesis which are not concerned with assessments of (propositional) contents intended to be communicated, such as the ascertainment of the function which some argument possesses in a context. For substantiating the thesis that omission of raising relevant questions concerning behaviour is not an isolated phenomenon two examples will be employed: (1) A discussion concerning the exegesis of a crucial passage of Dignāga’s Pramāṇasamuccaya and the Pramāṇasamuccayavṛtti, (2) a critical appraisal of a recent publication dealing with the interpretation of the second chapter of Nāgārjuna’s Mūlamadhyamakakārikā-s.

On the Inscrutability of Logic in Certain Natural Language Contexts

The paper opens by defining 'logical universality' as the retention of the propositional content of expressions under any enunciative circumstances. Universality in this sense, the paper claims, cannot be demonstrated in the same manner across different discursive domains and sign systems. Unlike in geometry, arithmetic, algebraic and mathematical logic, where logical universality can be shown to be non-controversial, the concept of universality becomes problematic as soon as natural language terms and syntax are employed. The paper shows the main reasons for this difficulty to lie in the extensional features of natural language, which cannot be adequately captured by intentional means. Intentional descriptions are claimed to apply only to semiotically homogeneous sign systems of a formal kind. Natural language expressions, in contrast, are semiotically heterogeneous, or heterosemiotic, characterised as they are by quasi-perceptual ingredients. Nevertheless, the paper argues, there are three cases in which logical universality can be demonstrated to hold in spite of natural language being employed, one of which is strictly technical language. In contrast, culturally fully saturated natural language use is shown to escape the constraints of logical universality as defined, on the grounds that some of its essential features, such as referential background, reference, and deixis, especially in its implicit form, effectively undermine the retention of identical propositional contents across cultures and time.