The system of rules that governs how we assign meaning to the morphemes we use is called (3 points)

1.1 Complexity and the Geisteswissenschaften: structuralism

For roughly the past half century, humanistic approaches to sociocultural anthropology have been dominated by the structural anthropology of Claude LéviStrauss, and the “post-structuralism” of his many successors, among them the Tel Quel group in Paris (1960–1982) which included Roland Barthes, Georges Bataille, Maurice Blanchot, Jacques Derrida, Michel Foucault and Julia Kristeva. Structuralism posed a profound challenge to the earlier humanistic tradition in anthropology, which sought to uncover the subjective meaning of cultural symbols and practices. Structuralists did away with the question of the subject's awareness of meaning, replacing it with an account of how language produces meanings that define subjects. The prominent structuralist Roland Barthes (1915–80) argued that the implications of this epistemological reversal could hardly be exaggerated, predicting that the “infinity of language” would replace the Kantian-Husserlian “infinity of consciousness.” The ascendancy of structuralism in anthropology in the 1960s created an ongoing philosophical crisis with respect to the nature of the anthropological subject, which continues today.indexmorphemes

Interestingly, it is probably easier to give a coherent account of the structuralist program from the perspective of complexity, than from that of humanistic anthropology. Structuralism defines various components of language, such as phonemes and morphemes, in terms of logical operations on trees or networks. This marked a radical departure from traditional interpretive approaches to language and culture. A century ago, the Swiss linguist Ferdinand de Saussure (1857–1913) defined the linguistic sign as comprised of two elements, the sensible sound-image (signifier) and the intelligible concept (signified). Saussure argued that linguistic signs are unmotivated and acquire their meaning only through differential relations with other signs.1 He suggested that the same dynamics occur at the level of phonology: the boundaries of phonemes are defined by paired contrasts with the other phonemes that they most closely resemble. Thus in English the slight difference between [p] and [b] marks the boundary between two phonemes, creating a meaningful distinction between, for example, “pit” and “bit.” In this way, binary contrasts or antonymy define signifiers: the written phonetic symbol [b] points to a particular sound (or range of sounds produced by different speakers). Roland Barthes later described this as first-order signification; i.e. the denotative meaning of the signifier. Barthes developed a concept of higher-order signifiers which enabled him to extend the structuralist approach from language to cultural phenomena. For example, the denotative or first-order meaning of the English signifier “blue” depends on the other color terms with which it can be contrasted. Barthes argued that second-order meanings are also defined by binary contrasts. Thus blue is traditionally associated with male infants, and pink with female infants, in American hospitals. This association is an example of metonymy: blue is to pink as male is to female. Barthes argued that such metonymic associations are ubiquitous, generating symbolic classificatory systems for cultural objects.

This idea was further developed by anthropologists such as Marshall Sahlins, who used it to analyze the systemic properties of cultural symbols. For example, Sahlins argued that the Fijian words for “sea” and “land” are first-order signifiers defined by their binary opposition: that which is sea is not land [Sahlins, 1976]. This contrast is extended by metonymic chaining: in Fiji, men are associated with the sea and women with the land; further, chiefs are also associated with the sea and commoners with the land. The seaward side of a Fijian house thus is associated with male and chiefly power. Similarly, the sea itself is subclassed into the lagoon (landward sea) and the outer or seawards sea. Fishing is a male occupation, but if women fish, they do so in the lagoon.

In this example, a relationship of binary opposition between two first-order signifiers “sea” and “land”, forms the root of a tree of symbolic associations (Fig. 1) in which the initial defining contrast is repeated with other paired oppositions, like seawards land and inland land.

The tree model was criticized by post-structuralists, who argued that there are no privileged first-order signifiers which unambiguously root trees of symbolic associations (thus the Sea/Land opposition in Sahlin's example would not be accepted as foundational). According to this argument, signification is not fully defined by any single oppositional pair in the chain of signifiers, but rather by metonymic associations. The post-structuralist psychoanalyst Jacques Lacan argued that the mind glides like a butterfly through networks of signifiers, each of which points beyond itself to other signifiers. Hence the correct model is not a rooted tree, but rather a network of signifiers: the chain of differences spreads throughout semantic space and never comes to rest in an ultimate ‘signified’ [Sandwell, 1996, 365–6]. This argument was elaborated by Jacques Derrida, who drew attention to the ‘free play’ of signifiers: they are not fixed to their signifieds but point beyond themselves in an ‘indefinite referral of signifier to signified’ [Derrida, 1978, 25]. Hence both Derrida and Lacan portray relationships among signifiers as networks, not trees. While for Saussure the meaning of signs derived from how they differ from each other, Derrida coined the term différance to allude to the ways in which meaning is endlessly deferred. He concluded that there is no ‘transcendent signified’ [Derrida, 1978,278–280; 1976, 20].

Derrida and other post-structuralists famously developed these ideas into a relativistic epistemology, arguing that the meaning of texts can never be fixed. This conclusion echoed that of many Anglo-American analytic philosophers, who at about the same time (1970s) had begun to acknowledge that their quest for an unambiguous observation language had failed. Meanwhile in linguistics, the structuralist program sputtered to an end as it became clear that networks defined by binary oppositions are not very informative for linguistic phenomena more complex than phonemes and morphemes.

2.5.4 Big Data and Parsing

Deep parsing of natural language text holds key to harnessing unstructured Big Data. For example, dependency tree (DT) is a representation produced via deep parsing. A DT is an acyclic graph and depicts dependencies between lexical entities (words or morphemes). When two words are connected by a dependency relation, one of the words is the head and the other is the dependent. A dependency link is an arrow pointing from the head to the dependent. Usually, the dependent is the modifier, and the head plays a larger role in determining the behavior of the pair in the text. There is growing body of work on creating new tree-banks for training dependency parsers for different languages.

The Nature of Human Language

Human languages are distinct from animal communication systems in a wide variety of ways. Among them are infinite creativity (the ability for speakers and hearers to produce and understand an infinite variety of utterances), their symbolic nature (the arbitrary relationships among words, utterances, and the concepts to which they refer), and hierarchical organization, which allows a number of levels of rules governing appropriate structure and use (see Fromkin et al., 2007).

Within any language, there is a set of rules that governs appropriate use of sounds, words, grammar, and meaning. Moreover, competent language users must also master socially appropriate means of conveying and interpreting linguistic messages. Briefly, these subsystems of language knowledge consist of phonology, morphology, syntax, semantics, and pragmatics.

Phonological features of any given language specify its sound inventory (phonemes) as well as ways in which sounds may be legally combined to create well-formed words (phonotactics). The smallest units of language that convey meaning or grammatical distinctions are morphemes; for example, in English, a word such as cats consists of one lexical morpheme, cat, which can stand alone, and one grammatical morpheme to signal the plural. Languages have large numbers of lexical (or open class) morphemes, and a much smaller and delimited number of grammatical morphemes (closed class), which may stand alone (such as the or can), or must be attached to lexical morphemes (such as the plural, past tense, possessive, etc.)

Mastery of syntax requires appropriate use of morphology as well as any rules governing the ordering of elements in sentences and their smaller constituents, such as noun and verb phrases. Some languages, such as Finnish, permit fairly free word order, while others are highly constrained. In addition, languages may differ in basic word order; for example, English tends to employ subject–verb–object as its canonical ordering, while the Philippine language Tagalog is primarily verb–subject–object.

Meaning in language can be conveyed by the meanings of individual words (as in knowing what the word chair refers to), as well as the order in which words are combined to reflect themes such as subject and object (most readers will readily appreciate that “John loves Mary” does not necessarily mean that “Mary loves John”). Finally, the meanings of sentences often go beyond the strict interpretation of their words and syntax. Pragmatic intent is obtained by evaluation of the sentence within a context to ascertain its function within conversation – whether one's objective is to inform, warn, request action, etc., as might variously be the case in hearing someone say, “It's late.” All of these rule systems must be adequately mastered in order for the child to function as a capable speaker–hearer of a language.

Vocabulary

A student’s understanding of word meanings, or their vocabulary, is often considered the glue that binds word recognition and comprehension. For students with special needs, gaps in vocabulary development cause reading delay and affect reading growth (Bos and Vaughn, 2006). There are two types of vocabulary instruction, oral vocabulary and reading vocabulary. Instruction in reading vocabulary (the words a reader recognizes in print or uses in writing) is the focus of this synopsis. Although teacher-directed instructional approaches such as oral language instruction (i.e., teaching vocabulary during reading by modeling how to use context; using synonyms and definitions) work well for average achieving students, students with special needs often require additional explicit instructional approaches.

DI for word-learning strategies, morphemic and contextual analysis, is an approach often used to support student independence in developing vocabulary. In morphemic analysis, students are taught to use word parts to interpret word meanings. For example, the teacher introduces and teaches the meaning of a morpheme (e.g., Pre usually means before, what does pre mean?). The teacher then teaches new words that include the morpheme through signaling and unison oral responding (e.g., So what does pretest mean?). Carnine et al. (2004) cautioned that this approach is limited by the difficulties in translating some words into functional definitions, the fact that many words have dual meanings, and the difficulty in selecting appropriate morphemes for instruction. Contextual analysis, in contrast, fosters students’ independence in learning word meanings by using surrounding words in text. In contextual analysis instruction, the teacher points out the unknown word in the text, asks the students to find the words that tell what the word means, and prompts the student to restate the sentence substituting the known word for the unknown word. Students are taught to look in text for an imbedded word-definition, a synonym, a description, a contrast, or a comparison.

Mnemonics with modified CSI has been used effectively for teaching vocabulary and associated concepts to students with special needs (Hughes, 1996). The keyword picture strategy, for example, uses visualization in the following steps taught to students: (1) select a word or term; (2) state the definition; (3) select another similar sounding word; (4) create a picture using the definition, the word to be learned, and the key word; and (5) think about and study the picture. When developing mnemonic strategies for vocabularies and concepts, Mastropieri and Scruggs (1989) recommended that the amount of elaboration required is dependent on the students’ familiarity with both the word and the concept abstraction.

Essential Underpinnings of Reading

All beginning readers depend on language underpinnings that begin developing in the preschool years. During formal reading instruction, children continue to develop oral-language capabilities needed to approach and understand written language. As preschool children's oral language develops, their expressive oral language increases and their listening comprehension becomes more sophisticated. As this occurs, their ability to remember and use information increases and they learn how to listen differently when hearing expository text, storybooks, poems, and nonsense rhymes. They become aware of the sounds of the language and the phonemes and morphemes that comprise words. They learn the purpose of written language and the forms it takes. At a basic level, they learn how print works (concepts of print). A final but major part of an early literacy foundation is children's developing motivation to read. These essential skills and knowledge continue to develop during instruction in learning to read and as students continue to build reading competence (see Figure 1). Below, we elaborate on these underlying skills and knowledge.

Issues in Assessing Language Acquisition

9.5.1 The model

An overview of the model is shown in Figure 9.3. It combines the GG model with a Boolean gene regulation network in each cell and an evolutionary process to focus on networks which generate morphogenesis. (For a detailed description of the model see Hogeweg (2000a).) Development starts with one large cell that undergoes a number of prescheduled cleavages; during the first two cleavages ‘maternal factors’ may cause cell differentiation. The rules for subsequent cell growth and division are based on the experimental observation that stretch can indeed lead to cell growth and division and squeezing of cells leads to apoptosis (Chen et al., 1997; Ruoslahti, 1997). Note that both these processes involve an elaborate sequence of changes in gene expression; we here implement only the effect. Cell division occurs in the model when a cell has reached a size that is twice some reference size and cell division is through the middle and perpendicular to the longest axis. This seems indeed to be the default situation and modelling of the spindle formation has shown that it will self-organize accordingly (Dogterom et al., 1995). However, in vivo many mechanisms are used to alter the division plane. We will examine below their effect on developmental robustness.

In earlier work (Hogeweg, 2000a, b, 2002) we have shown:

Differentiation into cell types varies between (a) stable differentiation, i.e. differentiation is maintained independent of intercellular signalling, (b) history dependent differentiation, in which the cell type is dependent on a sequence of neighbourhood conditions, and (c) differentiation which is fully specified by the current neighbourhood only. Each of these types of differentiation is associated with different types of morphogenesis.

Morphogenesis results as ‘sustained transient’ from surface energy minimization and ‘intrinsic conflict’, which is maintained by cell differentiation, cell growth and cell death. Without continued ‘interference’, the initial, high energy state would change, through shape changes to, at the end a ‘blob’-like low-energy shape.

These intrinsic conflicts lead to automatic orchestration of adhesion, migration, differentiation, cell growth/division and death. It results in ‘pseudo-isomorphic outgrowth’. Although, the shapes do change during ‘maturation’, a ‘critter’ preserves its general appearance.

Many different morphemes result from combinations of a few mechanisms:

An important mechanism is meristematic growth, a layer of dividing cells that differentiate into non-dividing (or rarely dividing) cells of several types. The zone is maintained because cells redifferentiate if ‘out of line’: cells differentiation that is fully dependent on neighbourhood is ‘used’.

A related mechanism we dubbed elongation by ‘budding’: a small group of differentiated cells is pushed outwards, because an other cell type on the one hand tends to engulf them, but on the other sticks together more firmly than to the ‘bud’. Again, the situation is maintained, because cells which do, nevertheless, engulf, differentiate into bud-type cells. Like the previous one, this mechanism depends on neighbourhood dependent cell differentiation. The elongation shown in Figure 9.4 is an example of this.

Another often occurring mechanism is convergence extension, which occurs due to maximization of the contact line between stably differentiated cell types and often involves redifferentiation of subtypes of these cell types when the contact zone increases.

Elongation can also result from intercalation of stably diverged cell types and their subsequent growth and division.

Finally engulfing is an intrinsic mechanism of differential cell adhesion. In our model it often induces neighbourhood dependent cell differentiation.

The evolutionary dynamics show many of the features known to result from a non-linear genotype-phenotype mapping, i.e. neutral paths and punctuated equilibria at the phenotypic level, although the shape of the quasispecies distribution differs from the simpler examples studied before (e.g. Huynen et al., 1996; van Nimwegen et al., 1999). We will examine this below in more detail.

Moreover, the evolutionary dynamics give rise to interesting mosaic-like variation at the phenotypic level, i.e. repeated reinvention of similar morphotypes occurs in one evolutionary history.

All these features involve the interaction of the intra-and intercellular levels, which generate long-range correlations. In the next two sections we will examine the relationship between the ‘informatic’ processes, i.e. gene regulation, and the dynamic processes, i.e. cell behaviour due to differential adhesion, by focusing on the morphological variation that occurs among critters with identical genomes. In other words, we will study how the genetic and inheritable information can on the one hand exploit and on the other hand ‘tame’ the dynamics of the system.

Aspectual coercion

Aspectual coercion, in which an aspectual operator is applied to a verb phrase denotation, which specifies an eventuality type inter alia, to produce another verb phrase denotation and eventuality type description, is another example of a meaning shift. Aspectual coercion is quite language specific, and thus is not the result of any general pragmatic operation such as that considered by Neo-Griceans or relevance theorists such as [Sperber and Wilson, 1986; Recanati, 2004]. Consider, for example, (6), which involves the progressive aspect.

#John is knowing French.

John is being silly.

John is just being John.

John's being an asshole.

One of the truisms about the progressive aspect is that stative constructions don't support it, as shown in (6a). Nevertheless, (6b-d), which are progressivizations of the stative constructions John is silly, John is John, and John is an asshole, are perfectly unproblematic. Interestingly, aspectual coercion with the progressive appears to be a particular feature of the English progressive aspect morpheme. Languages like French that lexicalize progressive aspect do not seem to support this meaning shift:

Jean est idiot.

#Jean est en train d'être idiot.

Jean est en train de faire l'idiot.

Aspectual coercion is a thus language specific phenomenon and thus cannot be the result of a general cognitive principle of strengthening or weakening due to Gricean or Neo-Gricean constraints on communication. Such meaning shifts must be a part of the linguistic system, due to the meaning of particular words.

Another language specific aspectual coercion concerns the application of a perfective aspectual operator to a verb phrase containing an ability modal. Consider the following French examples. (8) translates roughly as Jeanne had to take the train and (8a) and (9a) use the perfective aspect, while (8b) and (9b) have imperfective aspect.

Jeanne a du prendre le train. → Jeanne a pris le train.

(Jeanne had to take the train. → Jeanne took the train).

Jeanne devait prendre le train. ↛ Jeanne a pris le train.

(Jeanne was supposed to take the train. ↛ Jeanne took the train.)

(9)a.

Jeanne a pu prendre le train. → Jeanne a pris le train.

(Jeanne was able to take the train. → Jeanne took the train.)

Jeanne pouvait prendre le train. ↛ Jeanne a pris le train.

(Jeanne was able to take the train. → Jeanne took the train.)

The → signifies an actuality entailment. Were we to consider ability modals as true modals that we can symbolize with □ and ⋄, the actuality entailments in (8a) and (9a) would translate, respectively, to (10a) and (10b):

□ϕ → ϕ.

□ϕ → ϕ or ϕ → □ϕ.

which implies a collapse of the modality (Bhatt 1999). However, with the imperfective aspect, these inferences vanish, and there is no collapse. The puzzle is, how can an application of the aspectual perfect collapse the modality? This is unpredicted and indeed bizarre from a Montagovian view of composition.

Actuality entailments with certain verb forms, like coercion with the progressive aspect, is a phenomenon particular to certain languages. In English, for instance, the actuality entailment does not appear to exist:

John was able to take the train.

John had to take the train.

?John has been able to take the train.

None of these have the actuality entailment, though they might have what one could call an actuality implicature. Once again, the actuality entailment cannot be the result of some general cognitive but non linguistic principle of strengthening. It is a semantic and lexically constrained kind of inference.

Matters are still more complex when one considers how temporal adverbials interact with modality and aspect to produce actuality entailments.5

Soudain, Jean pouvait ouvrir la porte.

(Suddenly, Jean could open the door.)

In (14) the actuality entailment holds, despite the fact that the imperfective aspect is used. This is explained by the general observation that adverbs like suddenly coerce the imperfective aspect into an incohative one with a perfective meaning. But once again we have a shift of meanings.

I believe that the apparent meaning shifts discussed in 1.2-1.2 should receive as uniform a treatment as possible within a semantic/pragmatic framework of lexical meanings and semantic composition—that is, how lexical meanings compose together to form meanings for larger semantic constituents like propositions or discourses. But we can only address this issue adequately within a larger view of how context affects interpretation. To this end, I will review the outlines of how dynamic semantic frameworks, including theories like SDRT, view discourse content computation. This will give us the tools with which to understand context effects at the level of clausal content composition and apparent meaning shifts. I will then discuss a couple of classic meaning shift cases and spell out the general approach to these that I favor, comparing it to recent pragmatic as well as semantic accounts.

3 Document Preprocessing

We begin this section by defining terminology. Indexing is the process of creating both a representation for documents and associated data structures for storing and retrieving the representation. It is this representation that is used to determine the relevance of a document to a user query. For example, an inverted index is one such representation (discussed in Section 6).

1 Introduction

Human languages are the archetypal examples of a code, taken in the sense of “a small set of arbitrary rules selected from a potentially unlimited number in order to ensure a specific correspondence between two independent worlds.” (Barbieri, 2015) The two independent worlds in this case are the world of speech, gesture or written marks on the one hand, and the world of meaning on the other. The set of rules of a language consists of its lexicon (associating words or morphemes with meanings and functions) and its grammar (prescribing how larger units are built and how the meaning of these combinations is assembled to form the meaning of the whole).

Human languages differ from other codes in important ways: (i) The rules are not static but change over time, possibly in profound ways. Speech sounds obviously change, new words come into the language all the time and others become obsolete, grammatical marking may erode and be re-invented again. (ii) The rules are not universal. There is a large variety of languages and dialects within languages. Languages differ in all respects including in the kinds of meanings they are able to express directly. (iii) The rules are not made up by designers (as would be the case for a programming language) but they emerge dynamically by the individual activities of speakers and listeners. Language conformity is an emergent phenomenon without central control. (iv) The set of rules is not small and not definable. A typical language user employs on the order of half a million lexical and grammatical constructions and this set continuously expands and contracts based on usage. (v) Language is an inferential code, which means that the decoder needs a lot of contextual and background knowledge in order to be able to understand what is being said. This contrasts with most other codes where all the information to be transmitted is explicitly coded in the message.

All these properties make human language a fascinating object of study for code biologists − who tend to emphasize the dynamic evolutionary nature of the linguistic code and enquire about its origins and incessant change, in contrast with linguistic theorists who tend to seek the common static structure underlying language.

This paper addresses two open questions: (i) How does language develop in the individual? and (ii) How do languages evolve in a population? s is here used in the sense of change historically over time, which may involve real novelty such as the emergence of phrase structure or the emergence of a grammatical case system. Both questions are clearly related to each other, because language development in each individual of a community drives the evolution of its common language. We argue that the perspective of evolutionary biology, in particular the framework of replicator dynamics, helps to tackle both questions.