Evidence for Frames from Human Language

Parts and boundaries

Matching two natural language sentences is a fundamental problem in both natural language processing and information retrieval. Preliminary studies have shown that the syntactic structures help improve the matching accuracy, and different syntactic structures in natural language are complementary to sentence semantic understanding. Ideally, a matching model would leverage all syntactic information. Existing models, however, are only able to combine limited (usually one) types of syntactic information due to the complex and heterogeneous nature of the syntactic information. To deal with the problem, we propose a novel matching model, which formulates sentence matching as a representation learning task on a syntactic-informed heterogeneous graph. The model, referred to as SIGN (Syntactic-Informed Graph Network), first constructs a heterogeneous matching graph based on the multiple syntactic structures of two input sentences. Then the graph attention network algorithm is applied to the matching graph to learn the high-level representations of the nodes. With the help of the graph learning framework, the multiple syntactic structures, as well as the word semantics, can be represented and interacted in the matching graph and therefore collectively enhance the matching accuracy. We conducted comprehensive experiments on three public datasets. The results demonstrate that SIGN outperforms the state of the art and also can discriminate the sentences in an interpretable way.

Recognition Textual Entailment is a fundamental research direction in natural language processing. However, due to the complexity of sentence structure and the multiple parts of speech of words, it affects the model's deep understanding of sentences. From a linguistic perspective, different parts of speech have different understandings of sentences. We believe that a complete sentence consists of two parts: sentence structure and syntactic structure. In response to this issue, this article proposes a recognition textual entailment model that integrates attention mechanism and part of speech tagging. NLTK (Natural Language Toolkit) is used to annotate the part of speech of sentences, and the part of speech information is interactively fused with the sentence features extracted by the Conformer model. After part of speech tagging, the model can better grasp the syntactic structure of sentences. Experiments have shown that the above method combines sentence structure and syntactic structure, which can effectively enhance the model's understanding of deep semantics and improve its accuracy compared to classical models.

Syntactic parsing deals with syntactic structure of a sentence. The word 'syntax' refers to the grammatical arrangement of words in a sentence and their relationship with each other. The objective of syntactic analysis is to find syntactic structure of a sentence which is usually depicted as a tree. Identifying the syntactic structure is useful in determining the meaning of a sentence. Natural language processing is an arena of computer science and linguistics, concerned with the dealings amongst computers and human languages. It processes the data through lexical analysis, Syntax analysis, Semantic analysis, Discourse processing, Pragmatic analysis. This paper gives various parsing methods. The algorithm in this paper splits the English sentences into parts using POS tagger, It identifies the type of sentence (Facts, active, passive etc.) and then parses these sentences using grammar rules of Natural language. The algorithm has been tested on real sentences of English and it accomplished an accuracy of 81%.

This chapter outlines key findings of cognitive and developmental psychology which could be used as a theoretical framework to guide the design and research of e-learning applications. The chapter consists of two main sections. The first section presents the basic cognitive mechanisms and their development, while the second part discusses how our knowledge of the cognitive system can guide the design of computer-based learning environments. It is proposed that the human mind is organized in three levels: two of them are general-purpose mechanisms and processes and one consists of domainspecific structures of knowledge representation and problem solving processes. These three levels are associated with an effective learning framework. The suggested framework describes the basic points that any designer of e-learning environments must consider. The second section of the chapter discusses e-learning, connecting it to knowledge about the human cognitive system. In this section, we first present the conceptual bounds of the e-learning construct and discuss how cognitive theories of learning should guide the design of technologically-enhanced learning environments. It is proposed that a synergy between the study of human cognition and the design of e-learning applications is required for effectively understanding both fields.

Polysemy has been considered in terms of sense relations where each sense of a polysemous expression is represented individually in the lexicon. Siswati and other African languages accounted for in African Wordnet do not differ in the characterisation of this phenomenon. Polysemy is relevant in linguistic debates because of its inherent lexical meaning representation and compositional semantics. Because polysemy proliferates in natural languages, it is distinguished from monosemy and homonymy, despite the fact that they may seem clear at an intuitive level. However, they have proven difficult to draw from practice to some extent. This article explores polysemy and its role in lexical sense relations in African languages, specifically Siswati. It further examines the inherent aspects of word meaning and the semantic relations between words, as well as the ways in which word meaning is related to syntactic structure, given that syntactic structure is the basis upon which the synsets and usage examples are formulated. The article further demonstrates aspects that make the Siswati language system susceptible to polysemy. It further considers the pragmatic approach in accounting for lexical semantics as this provides the basis for a unified account of the role of polysemy in numerous domains, and for the amplification that motivates its proliferation in natural languages.

We describe a deterministic shift-reduce parsing model that combines the advantages of connectionism with those of traditional symbolic models for parsing realistic sub-domains of natural language. It is a modular system that learns to annotate natural language texts with syntactic structure. The parser acquires its linguistic knowledge directly from pre-parsed sentence examples extracted from an annotated corpus. The connectionist modules enable the automatic learning of linguistic constraints and provide a distributed representation of linguistic information that exhibits tolerance to grammatical variation. The inputs and outputs of the connectionist modules represent symbolic information which can be easily manipulated and interpreted and provide the basis for organizing the parse. Performance is evaluated using labelled precision and recall. (For a test set of 4128 words, precision and recall of 75% and 69%, respectively, were achieved.) The work presented represents a significant step towards demonstrating that broad coverage parsing of natural language can be achieved with simple hybrid connectionist architectures which approximate shift-reduce parsing behaviours. Crucially, the model is adaptable to the grammatical framework of the training corpus used and so is not predisposed to a particular grammatical formalism.

In this paper we develop novel algorithmic ideas for building a natural language parser grounded upon the hypothesis of incrementality. Although widely accepted and experimentally supported under a cognitive perspective as a model of the human parser, the incrementality assumption has never been exploited for building automatic parsers of unconstrained real texts. The essentials of the hypothesis are that words are processed in a left-to-right fashion, and the syntactic structure is kept totally connected at each step. Our proposal relies on a machine learning technique for predicting the correctness of partial syntactic structures that are built during the parsing process. A recursive neural network architecture is employed for computing predictions after a training phase on examples drawn from a corpus of parsed sentences, the Penn Treebank. Our results indicate the viability of the approach and lay out the premises for a novel generation of algorithms for natural language processing which more closely model human parsing. These algorithms may prove very useful in the development of efficient parsers.

In the era of digitalization, the interaction between humans and machines, particularly in Natural Language Processing, has gained crucial importance. This study focuses on improving the effectiveness and accuracy of chatbots based on Natural Language Processing. Challenges such as the variability of human language and high user expectations are addressed, analyzing critical aspects such as grammatical structure, keywords, and contextual factors, with a particular emphasis on syntactic structure. An optimized chatbot model that considers explicit content and the user’s underlying context and intentions is proposed using machine learning techniques. This approach reveals that specific features, such as syntactic structure and keywords, are critical to the accuracy of chatbots. The results show that the proposed model adapts to different linguistic contexts and offers coherent and relevant answers in real-world situations. Furthermore, user satisfaction with this advanced model exceeds traditional models, aligning with expectations of more natural and humanized interactions. This study demonstrates the feasibility of improving chatbot–user interaction through advanced syntactic analysis. It highlights the need for continued research and development in this field to achieve significant advances in human–computer interaction.

How speakers sequence words and phrases remains a central question in cognitive psychology. Here we focused on understanding the representations and processes that underlie structural priming, the speaker’s tendency to repeat sentence structures encountered earlier. Verb repetition from the prime to the target led to a stronger tendency to produce locative variants of the spray-load alternation following locative primes (e.g., load the boxes into the van) than following with primes (e.g., load the van with the boxes). These structural variants had the same constituent structure, ruling out abstract syntactic structure as the source of the verb boost effect. Furthermore, using cleft constructions (e.g., What the assistant loaded into the lift was the equipment), we found that the thematic role order (thematic role-position mappings) of the prime can persist separately from its argument structure (thematic role-syntactic function mappings). Moreover, both priming effects were enhanced by verb repetition and interacted with each other when the construction of the prime was also repeated in the target. These findings are incompatible with the traditional staged model of grammatical encoding, which postulates the independence of abstract syntax from thematic role information. We propose the interactive structure-building account, according to which speakers build a sentence structure by choosing a thematic role order and argument structure interactively based on their prior co-occurrence together with other structurally relevant information such as verbs and constructions.

Evolving pragmatics

In this paper, we show how a constraint-based approach influences the modeling of (1) preposition lexicalization and of (2) the choice of syntactic structure in natural language generation (in particuler in French). We concentrate on the linguistic description, which is the most challenging. The CSP procedures themselves are then rather straightforward. In the first case, preposition choice depends on the verb and its requirements, on the one hand, and the characteristics of the NP the preposition heads, on the other hand. In the second case, the choice of a particular syntactic structure depends on syntactic, semantic and pragmatic constraints on a verb, its arguments and on the parameters of the semantic representation. These dependencies may induce somewhat contradictory expectations. A constraint-based approach introduces a declarative model of these complex relations, allowing to identify all the possible (1) lexical choices or (2) syntactic structures which can be predicted from lexical descriptions.KeywordsSyntactic StructureConstraint Satisfaction ProblemNatural Language GenerationLinguistic DescriptionMovement VerbThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

The applications of using Synchronous Computer-Mediated Communication (SCMC) tools, known as real time text-message exchange software such as MSN Messenger, chat rooms, etc., in second language learning classrooms have been examined by numbers of researchers in the past few decades (e.g. Beauvois, 1992, 1998; Chun, 1994; Kern, 1995; Lai & Zhao, 2006; Sykes, 2005; Vandergriff, 2006). Motivated by an interest in the transferability from actually “typing” to “speaking” brought up by Abrams (2003), and the strong recommendations of using online discussion as a “stepping stone” in communication, the present study examined oral production in terms of accuracy, lexical complexity, syntactic complexity, and fluency under two modes of discussions in an EFL freshman English course. Subjects were 20 English-major students enrolled in a required university Freshman English class in an Asian EFL context. Data were collected under two modes: Synchronous Computer-Mediated Communication (SCMC) and Face-to-Face (FTF). Under each mode, students participated in two activities: (1) preparation discussions for an informal debate, and (2) discussions of issues after video watching. Each discussion lasted for ten minutes and was followed by an immediate FTF oral discussion. In FTF discussions, the audio was recorded and transcribed for further analysis. Meanwhile, the discussions in the SCMC context were collected as chat logs. All the collected data were divided into four parts: (1) discourse in traditional face-to-face discussion and (2) its follow-up discussion; (3) discourse in the SCMC context and (4) its follow-up oral discussion. The results revealed significant differences among all variables in comparing the language outputs in SCMC and FTF discussions. In fact, language outputs in SCMC exceeded the outputs in FTF in almost every aspect. The findings revealed lower error rate, fewer dysfluent markers, and higher percentages of using more sophisticated words in SCMC. Only the complexity of syntactic structures was greater in FTF. We also found positive effects of language production on all variables in FTF discussion activities; however, only fluency showed significant improvement. In terms of transferability from SCMC to oral proficiency, none of the variables showed significant effects on its immediate follow-up oral discussion. Ironically, the differences between SCMC and its follow-up FTF discussion were all significant except for syntactic complexity. This finding indicated that learners were not able to carry the “refined” utterances in SCMC over to oral discussion, or they just inevitably made the same mistakes when the communication channel switched back to normal FTF. Finally, we compared the effects of practicing in FTF modality and the effects of practicing in SCMC modality. The findings indicated that learners can produce more fluent and more complex utterances orally by practicing in FTF all the way through, while accuracy and lexical complexity can be increased by practicing in the SCMC environment. In conclusion, transferability can only be found partially in terms of accuracy and lexical complexity. Though the effects of practicing through SCMC with the aim of improving oral proficiency did not differ significantly from traditional face-to-face practices, it still shed some light on the potential benefits of text-based communication. With less time pressure through chatting in SCMC, learners were encouraged to retrieve from their long term memories and use less frequent words so that the vocabulary which was retrieved could easily be used again in later speaking activities.

Rapid development of natural language processing technologies has paved way for automatic sentiment analysis and emergence of robo-readers in computational finance. However, the technology is still in its nascent state. Distilling sentiment information from unstructured sources has turned out to be a complicated and strongly domain-dependent problem. To emulate the human ability to recognize financial sentiments in natural language by using machines, we need to provide them with (i) necessary ontological knowledge on the relevant domain-concepts, and (ii) learning strategies that help the machines to combine this knowledge with the syntactic structures extracted from text. In this paper, we present a knowledge-driven tree kernel framework for sentence-level analysis of financial news sentiments. Comparisons with linear kernels and classical lexicon-based systems suggest that significant performance gains can be achieved by incorporating information on financial concepts and their grammatical context. The framework is decomposable into learning, knowledge and syntactic structure components. Contribution of each part is separately examined using a human-annotated phrase-bank with close to 5000 sentences collected across a number of financial news sources. The proposed sentiment analysis framework is flexible and can be applied also outside financial domain. To evaluate cross-domain performance, a further comparison of the algorithms is done with datasets from non-financial domains including movie reviews and general political discussions.

The defining property of a Concept-to-Speech system is that it combines language and speech generation. Language generation converts the input concepts into natural language, which speech generation subsequently transforms into speech. Potentially, this leads to a more ‘natural sounding’ output than can be achieved in a plain Text-to-Speech system, since the correct placement of pitch accents and intonational boundaries —an important factor contributing to the ‘naturalness’ of the generated speech— is co-determined by syntactic and discourse information, which is typically available in the language generation module. In this paper, a generic algorithm for the generation of coherent spoken monologues is discussed, called D2S. Language generation is done by a module called LGM which is based on TAG-like syntactic structures with open slots, combined with conditions which determine when the syntactic structure can be used properly. A speech generation module (SGM) converts the output of the LGM into speech using either phrase-concatenation or diphone-synthesis.