連体形 stem + き stem + かる stem+しき stem + しかる 已然形 stem + けれ - stem + しけれ -
命令形 - stem + かれ - stem + しかれ
This is a lot of inﬂectional potential‚ but as classical Japanese transitioned to modern Japanese‚ all these forms have essentially become merged‚ leading to a single inﬂectional scheme that mixes forms from the ‘pure’ versions of adjectives with the ある-contracted versions of those adjectives‚ leading to the question of which forms are to be considered belonging to the adjective as it exists now‚ and which belong to the the verb ある‚ which happens to work together with verbal adjectives a lot. In this book‚ we’ll consider the ﬁnal inﬂected 命令形 for verbal adjectives to be a contraction of the verbal adjective’s 連用形 and the 命令形 for the verb ある‚ which is あれ. This gives us “verbal adjective stem + く” + “あれ” → “verbal adjective stem + くあれ”‚ where くあ contracts to か‚ giving us a ﬁnal rule “stem + かれ”. So‚ in this book‚ verbal adjectives are considered not to have a genuine 命令形 of their own‚ instead relying on the helper verb ある for one. However‚ other books list it as being simply “stem + かれ”‚ and so for completeness it has been included in the earlier table of bases.
Conventional sentence compression meth- ods employ a syntactic parser to compress a sentence without changing its mean- ing. However, the reference compres- sions made by humans do not always re- tain the syntactic structures of the original sentences. Moreover, for the goal of on- demand sentence compression, the time spent in the parsing stage is not negligi- ble. As an alternative to syntactic pars- ing, we propose a novel term weighting technique based on the positional infor- mation within the original sentence and a novel language model that combines statistics from the original sentence and a general corpus. Experiments that involve both human subjective evaluations and au- tomatic evaluations show that our method outperforms Hori’s method, a state-of-the- art conventional technique. Because our method does not use a syntactic parser, it is 4.3 times faster than Hori’s method. 1 Introduction
We have the possibility to introduce ambiguous readings in many cases and leave the disambiguation to a disambiguation module, or analyse all topics as modifiers to the sentence and leave the linking to a zero pronoun resolution module. In both cases, there is the necessity to rely on a natural language processing module that has access to a different type of information than the HPSG grammar processing. The introduction of ambiguity is useful when parsing not too long sentences and building up treebanks with interfering human evaluation, such as being done in the Hinoki project (Bond et al. 2004). The underspecification of information is useful when parsing large amounts of data containing long sentences and much topicalization, such as was done in the Verbmobil project (Wahlster 2000). As there are different demands for different kinds of processing, we decided to insert the possibility for ambiguous readings and set a switch to the root node of the grammar that constraints the application of the lexical entry which replaces the case particle if required. The topic particle gets three lexical entries. The first one is for the verb modifying topic variant, as in Example 168, Example 169 and Example 170. The second entry is for the case marking variant of wa, as in Example 167, where the case is assigned ga. It gets the same head as the other case marking particles and does not add to the semantics, just like case particles. In the case of a topic particle wa replacing wo, there is furthermore empathy set to the entity marked by the particle, as Watanabe (2000) states. This topic case particle as well does not add to the semantics, but to CONTEXT: it adds empathy setting to the entity it attaches to.
all the information (and constraints) of both (as long as it is not inconsistent). Thus we are building larger and larger information structures into the scientific models of the grammar. I think we have moved quite a distance from minimalist idealisation and the modelling practices that come with it. One aspect of minimalist ideali- sation is that de-idealisation is very often not possible, recall the Ising model of ferromagnetism mentioned in Sect. 4 in which particles are represented simply as points along a line. Feature structures and model-theoretic syntax allow for grad- ability of representation and the re-introduction of removed material. To be more specific, in formal language theory, sentences are modelled as semantically vacuous strings. A grammar, as a generative device, specifies the types of rules applica- ble to these strings in order to generate different sets of strings, i.e. languages, of varying complexity. For example adding recursive rules to the rules which gener- ate regular grammars gives rise to context-free grammars and so on (this picture is overly simplistic for the sake of illustration). Given this idealisation, there is simply no room for semantic content or phonological character to enter into the resulting model. These aspects of language are dealt with separately (in line with isolation idealisation). Contrasted with this, in constraint-based or model-theoretic approaches we can admit as much information into the base syntactic feature structures as we like, including semantic and phonological features. The models are not incompatible with introducing more and more information or features in order to “come closer” to the real world target system. This is the hallmark of Galilean idealisation (or mere abstraction) as (Weisberg 2007, 2013) describes it or distortion for the sake of computational tractability with the possibility of reintroduction of abstracted or idealised material. However, this is generally not a feature of minimalist idealisa- tion.
Transformational Grammar is a version of a larger set of different versions of Generative Grammar. Generative Grammar developed in the 1950s in the context of what came to be known as ‘the cognitive revolution’, which marked a shift to focusing on the mental processes underlying human behaviour from a mere concern with human behaviour for its own sake. As far as language is concerned, it marked a shift from a concern with the mechanics of certain limited aspects of language (mostly, morphophonemics) to a concern with the mental processes underlying a broader range of the properties of language. This change led to the articulation of certain ideas about the mental processes underlying language, some of which have been mentioned in the previous sections. Here we will limit ourselves to a brief and broad description of the evolution of some of the major ideas which have influenced the development of Transformational Grammar. Inevitably, some of the specialised terminology will not be transparent to the uninitiated reader, but, hopefully, will become so in the course of reading this book.
Clahsen's strategies make reliable predictions for the acquisition of word order. However, there are many points of criticism (cf. Pienemann, Johnston and Brindley 1987, White 1989, Eubank 1991). A major objection is that the concept of strategies was based on transformational grammar, which has since been abandoned by large groups of the linguistic community. It is also problematic that there is no explanation for how the constrained grammar develops in the first place, and that, as White (1989) points out, the definition of strategies partially relies on findings on comprehension, whereas Clahsen's model makes claims about speech production. Lastly, because the strategies are set up to prevent the movement of subconstituents across the boundaries of major constituents, the strategies approach is restricted to the phenomena of word order. However, the predictions of Clahsen's strategies are "rock solid" (Pienemann in press, 67), and the concept of constituent structures was found to be psychologically plausible. On these grounds, Pienemann later developed a model of processing constraints which are not based on constituent movements, but on transfer of abstract grammatical information across constituent boundaries (see ch. 1.4.2 below).
Recently, automatic generation of collocations for computational lexicography and online language learning has drawn much attention. Sketch Engine (Kilgarriff et al., 2004) summarizes a word’s grammatical and collocation behavior, while JustTheWord clusters the co-occurring words of single-word queries and TANGO (Jian et al., 2004) accommodates cross-lingual collocation searches. Moreover, Cheng et al. (2006) describe how to retrieve mutually expected words using concgrams. In contrast, GRASP, going one step further, automatically computes and displays the information that reveals the regularities of the contexts of user queries in terms of grammar patterns.
The DS framework is used to model postposing in several languages: English (Cann et al. 2004), Greek (Chatzikyriakidis 2011, Gregoromichelaki 2013), and Mandarin (Wu 2005). These studies are primarily concerned with NP postposing, but our data confirm that a wider range of syntactic items may be postposed in Japanese. In this section, we propose a DS account of Japanese postposing by advancing formal aspects of the framework. For brevity, the analysis is based on artificial examples which preserve the essence of the narrative data.
This paper discusses findings of a frame-semantic contrastive text analysis of English and Japanese, using the large-scale and precise descriptions of semantic frames provided by the FrameNet project (Baker, 2006; Fillmore, 2006; Fontenelle, 2003) 1 . FrameNet is a lexicon-building project, which has been analyzing meanings of English lexical units in terms of semantic frames they evoke. It annotates corpus example sentences with frame-semantic analyses and incorporates them into the lexicon. This paper points out that even though the FrameNet methodology allows us to compare languages at a more detailed level than previous studies, in order to investigate how different languages encode the same events, it is necessary for the frame-semantic lexicon to specify grammatical affordances of its entries. Based on a contrastive text analysis of an English-Japanese aligned parallel corpus and on the lexicon-building project of Japanese FrameNet (Ohara et al., 2006), the paper attempts to represent interactions between lexical units and constructions of Japanese sentences in terms of the combined lexicon and “constructicon,” currently being developed in FrameNet (Fillmore, 2006).
Because XDG allows us to write grammars with completely free word order, XDG solving is an NP - complete problem (Koller and Striegnitz, 2002). This means that the worst-case complexity of the solver is exponential, but the average-case complex- ity for the hand-crafted grammars we experimented with is often better than this result suggests. We hope there are useful fragments of XDG that would guarantee polynomial worst-case complexity. 3 A Relational Syntax-Semantics Interface Now that we have the formal and processing frame- works in place, we can define a relational syntax- semantics interface for XDG . We will first show how we encode semantics within the XDG frame- work. Then we will present an example grammar (including some principle definitions), and finally go through an example that shows how the rela- tionality of the interface, combined with the con- currency of the constraint solver, supports the flow of information between different dimensions. 3.1 Representing Meaning
According to Hauser et al., “it seems relatively clear, after nearly a century of intensive research on animal communication, that no species other than humans has a comparable capacity to recombine meaningful units into an unlimited variety of larger structures, each differing systematically in meaning” (2002: 1576). Certainly, the communicative systems can be found among various animal species, but something as intricate as language has remained unobservable outside the human species. Therefore, highlighting the absence of recursion in the animals’ communicative systems is justified and yet redundant, just as it would be redundant to claim that 3D technology will not be available to the viewer via his colourless television. Moreover, it is not clear how the examination of such abstract concept like recursion and infiniteness among animals can be productive for our discussion. The comprehension of the term requires at least human cognitive capacities, which tells us that the application of the same would require similar capacities. In addition, the property of boundlessness expands beyond the realms of language. It is observed and discussed in most natural sciences: chemistry, physics, mathematics etc. Thus, even if the infiniteness is linguistically observable, there is no reason why it should the separating factor between the so-called LAD and other general cognitive capacities. Pinker and Jackendoff have criticized most of the argumentation regarding the recursion provided by Hauser et al. (2002), especially it being exclusively tied to FLN and the evidence of its existence (Pinker and Jackendoff 2005). The recursion can be observed or interpreted in other human senses, such as human visual cognition. As they explain, the outside world is perceived as being made of discrete elements, which can be joined together to form larger constituents, and the sequences which are observed as pairs or clusters can be endless. It is always possible to generate larger constituents from various elements (ibid, 2005). Since the universe itself is infinite and humans tend to categorize, it can be expected that the so called “discrete infinity” is perceivable wherever. If we were to gracefully integrate the property of recursion into the nativist approach, we ought to attribute the property to FLB rather than the FLN.
Various functional relations are expressed by particles in Japanese. For instance, particles such as bakari, dake, nomi specify focus in sentences. Focus particles bear diﬀerent syntactic functions depending on where they appear in the sentence, so a Japanese parsing system needs to be able to correctly treat these particles.
In every language, a verb describes an action which relates to a number of participants. Chapter 6 sets out 'The semantic basis for a typology', showing that in some languages the meanings of verbs are oriented towards type of participant (for example, 'eat meat' or 'eat vegetables') whereas in others they relate to type of action Ceat where a good deal of chewing is involved', 'eat by sucking', etc.). For every language, a set of open word classes can be recognised on language internal criteria (typically: noun, verb and adjective). Languages vary as to the number and types of ways they have for deriving a stem of one class on the basis of a form from another word class. Chapter 7 deals with 'Word-class changing derivations in typological perspective'.
Parsing spoken language without syntax Parsing s p o k e n language without syntax Jean Yves Antoine C L I P S I M A G B P 53 F 3 8 0 4 0 G R E N O B L E C e d e x 9, F R A N C E J e a n Y v e s A n t[.]
As with subcategorization (section 3), we need to make precise the basis for the linguistic categories being used. Since all learners are learning the same L2, there are common aspects to their interlanguage development, in spite of the potential influence of the native language (L1) (Ellis, 2008). We thus use the L2 as a reference frame for the annotation, to define properties such as: “this verb requires a nominative subject to the left.” While one might want to directly encode IL, it is not clear what terms like “subject” mean in such a case. Additionally, annotation reliability would be an issue for L1-based or IL-based annotation, as the same sentence can have different analyses depending upon the L1 (Gass and Selinker, 2008, p. 106).
Baker, Sidney J. 1945. The Australian language: an examination of tile English lAn guage and English speedl as used in Australia, from convict days to the present, witlt special reference to the growth of indigenous idiom altd its use by Australian writers. Sydney: Angus and Robertson.
(Goodman 1996) In supporting this claim, Sharon Goodman reminds her readers of the rich variety and presence of multimodal texts. Newspapers contain photographs, diagrams and changes of typeface. Even a company letterhead will be carefully designed, including the choice of graphics and colour of the paper to craft the company’s image. We now take it for granted that an electronic text, such as a page on the web, will use more than one of the language modes.
Tree based translation models are a com- pelling means of integrating linguistic in- formation into machine translation. Syn- tax can inform lexical selection and re- ordering choices and thereby improve translation quality. Research to date has focussed primarily on decoding with such models, but less on the difficult problem of inducing the bilingual grammar from data. We propose a generative Bayesian model of tree-to-string translation which induces grammars that are both smaller and pro- duce better translations than the previous heuristic two-stage approach which em- ploys a separate word alignment step. 1 Introduction
A SYNTAX PARSER BASED ON THE CASE DEPENDENCY GRAMMAR AND ITS EFFICIENCY A SYNTAX PARSER BASED ON THE CASE DEPENDENCY GRAMMAR AND ITS EFFICIENCY Toru Hitaka and Sho Yoshida Department of Electronics, K[.]