Search

The purpose of the article is to examine the Hebrew (im)politeness marker bevakasha ‘please’ (the prefixed preposition bϵ ‘in’ + the noun bakasha ‘request’) in order to account for its present-day polysemy. The analysis of bevakasha, a hitherto neglected marker by both synchronic and diachronic investigations, is based in this article on corpora documenting the various stages of the Hebrew language. The findings illustrate a change in the contexts in which the marker has been used throughout history, i.e. from religious texts and letters of authoritative ruling to secular, casual contexts, from authoritative to egalitarian relationships, from first person to second- and third-person's benefits and further to argumentative contexts. The article concludes that the change in contexts has led to pragmatic expansion and a reanalysis of the marker which gave rise to the meaning of (im)politeness and to a shift in the power balance between the speaker and the addressee. Furthermore, the study suggests that the grammaticalization path of bevakasha has been motivated by a changing orientation towards the various components of the speech act, i.e. speaker, hearer and discourse.

Semantic diversity refers to the degree of semantic variability in the contexts in which a particular word is used. We have previously proposed a method for measuring semantic diversity based on latent semantic analysis (LSA). In a recent paper, Cevoli et al. (2020) attempted to replicate our method and obtained different semantic diversity values. They suggested that this discrepancy occurred because they scaled their LSA vectors by their singular values, while we did not. Using their new results, they argued that semantic diversity is not related to ambiguity in word meaning, as we originally proposed. In this reply, we demonstrate that the use of unscaled vectors provides better fits to human semantic judgements than scaled ones. Thus we argue that our original semantic diversity measure should be preferred over the Cevoli et al. version. We replicate Cevoli et al.’s analysis using the original semantic diversity measure and find (a) our original measure is a better predictor of word recognition latencies than the Cevoli et al. equivalent and (b) that, unlike Cevoli et al.’s measure, our semantic diversity is reliably associated with a measure of polysemy based on dictionary definitions. We conclude that the Hoffman et al. semantic diversity measure is better-suited to capturing the contextual variability among words and that words appearing in a more diverse set of contexts have more variable semantic representations. However, we found that homonyms did not have higher semantic diversity values than non-homonyms, suggesting that the measure does not capture this special case of ambiguity.

There is a growing interest in the phenomenon of co-predication, where a single NP appears with predicates that have incompatible selectional preferences. The phenomenon suggests that a single NP can stand for at least two entities and so that the truth-conditions of a co-predicative sentence are more complicated than what the surface structure indicates. In this paper, two different kinds of proposals concerning this issue are presented and discussed: conservative approaches that commit to there being just one entity that is referred by a co-predicative term; and non-conservative, but more intuitive approaches, that assign conjunctions of simple truth-conditional contents to co-predicative sentences. Meanwhile, it is argued that, although much is still unknown about co-predication, many approaches coincide in signalling that the different senses of the terms that co-predicate in a stable way are linked by particular metaphysical relations.

This paper discusses methodological issues related to the implementation of experimental studies that investigate the processing of literal and metaphorical expressions. The first and central aim of the paper is to discuss methodological questions arising from current experimental research on figurative language, such as use of heterogeneous stimuli and their suitability for investigating research questions related to the processing of figurative language. At the center of the methodological discussion is the contextual and cognitive dominance of literal and metaphorical meanings in polysemous words as an important variable in experimental research on figurative language. In this paper, we discuss its importance and propose a possible way of controlling for this variable. The second aim of this paper is to present experimental results that complement the methodological discussion and show that metaphor processing in polysemous words (i.e. very conventionalized figurative meanings) is not necessarily more difficult or special in comparison to processing of literal language if the stimuli are properly balanced. In this paper, we suggest a way to design stimuli that are balanced both in terms of traditional lexical measures as well as dominance of the literal and metaphorical meanings retrieved from corpus data.

The word similarity task is used to calculate the similarity of any pair of words, and is a basic technology of natural language processing (NLP). The existing method is based on word embedding, which fails to capture polysemy and is greatly influenced by the quality of the corpus. In this paper, we propose a multi-prototype Chinese word representation model (MP-CWR) for word similarity based on synonym knowledge base, including knowledge representation module and word similarity module. For the first module, we propose a dual attention to combine semantic information for jointly learning word knowledge representation. The MP-CWR model utilizes the synonyms as prior knowledge to supplement the relationship between words, which is helpful to solve the challenge of semantic expression due to insufficient data. As for the word similarity module, we propose a multi-prototype representation for each word. Then we calculate and fuse the conceptual similarity of two words to obtain the final result. Finally, we verify the effectiveness of our model on three public data sets with other baseline models. In addition, the experiments also prove the stability and scalability of our MP-CWR model under different corpora.

This article proposes that predicate order and coherence relations are the two major determining factors in copredication licensing, resolving a long-standing puzzle over the criteria for construct...

Natural language inference (NLI) task requires an agent to determine the semantic relation between a premise sentence ( ${p}$ ) and a hypothesis sentence ( ${h}$ ), which demands sufficient understanding about sentences semantic. Due to the issues, such as polysemy, ambiguity, as well as fuzziness of sentences, intense sentence understanding is very challenging. To this end, in this article, we introduce the corresponding image of sentences as reference information for enhancing sentence semantic understanding and representing. Specifically, we first propose an image-enhanced multilevel sentence representation net (IEMLRN), that utilizes the image features from pretrained models for enhancing the sentence semantic understanding at different scales, i.e., lexical-level, phrase-level, and sentence-level. The proposed model advances the performance on NLI tasks by leveraging the pretrained global features of images. However, as these pretrained image features are optimized on specific image classification datasets, they may not show the best performance on NLI tasks. Therefore, we further propose to design an adaptive image feature generator that extracts fine-grained image labels from the corresponding sentences. After that, we extend the IEMLRN to multilevel image-enhanced sentence representation net (MIESR) by utilizing not only the coarse-grained pretrained features of images, but also the fine-grained adaptive features of images. Therefore, sentence semantic can be evaluated and enhanced more comprehensively and precisely. Extensive experiments on two benchmark datasets (SNLI, SICK) clearly show our proposed IEMLRN significantly outperform the state-of-the-art baselines, and our proposed MIESR model achieves the best performance by considering not only the text but also images in an adaptive multigranularities way.

The cognitive process and one of its most difficult aspects – translation into different languages - is an obstacle to creating a perfect Artificial Intelligence, because translation is not limited to formulas. The accuracy of the terminology and the lack of ambiguity of the word are the purposes of artificial intelligence while limiting the development of linguistic thinking. The transhumanism involves the improvement of the man, as an unfinished link of evolution, based on his connection with technological elements. From the author`s point of view, the “man-machine” interaction offers a dubious advantage for the human civilization and the planetary ecological system. This is, first of all, anachronistic and its main goal is singularity, i.e., the management of intellectual energy within the strict framework of technology. The article schematically provides the explicit sequence of the translation process. Based on relevant examples, the author demonstrates the difference between mental operations of Artificial Intelligence and those of the man in the translation process. According to the author, translation is a universal feature of human thought, while the process of translating the thought into the speech form is due to biological-genetic, emotional, socio-historical memory and experience, as well as the possibility of the individual translation of the thought into the mother tongue and into any other language. Moreover, translation is one of the strongest survival qualities of homo sapiens.

This paper focuses on the metaphor of the third person possessive pronoun ʦeǐ in Longxi Qiang, a Tibeto-Burman language. The discourse uses of ʦeǐ suggest that conceptual distance goes together with distance in surface realization. Speakers use ʦeǐ as a possessive pronoun to signal possession and definiteness. In addition to the referential meaning, speakers use ʦeǐ to express stance-taking, including the evaluation of spatio-temporal and relational distances, affective stance and discourse organization. In an utterance, ʦeǐ co-occurring with the filler lamo implicates that the speaker cannot think about how to say something about the third party. The polysemy of ʦeǐ is driven by the metaphorical transfer from possession to stance-taking. Possession can account for this metaphoric process.

Knowledge graph embedding aims to learn the embedded representation of entities and relations in knowledge graphs which is very important for the subsequent link prediction task. However, two key issues are existed for learning knowledge graph embedding: 1) How to take full advantage of the deep learning algorithms to generate expressive embeddings? 2) How to solve the polysemy phenomenon caused by multi-relations knowledge graphs that entities and relations show different semantics after involving different predictions? In this article, to tackle the first problem, the multi-layer convolutional networks are adopted to generate features about entities and relations then used to predict candidate entity. Moreover, the representation power of the networks is strengthened by integrating an effective recalibration mechanism which can accentuate informative features selectively. To tackle the second problem, we propose to learn multiple specific interaction embeddings. Instead of directly learning one general embedding to preserve all information for each entity and relation, their interactions are captured to model the cross-semantic influence from relations to entities and from entities to relations. Compared to traditional embedding models, the proposed model can provide more generalization capabilities and effectively capture potential links between entities and relations. Experimental results have revealed that the proposed model achieves the state-of-the-art performance for general evaluation metrics on link prediction tasks.

Considering that most popular models solving aspect-level sentiment classification problems mainly focus on designing complicated neural networks to scale the importance of each word in the sentence, this paper addresses this problem from the view of semantic space. Motivated by the fact that the senses of a word can be sophisticatedly embedded into the semantic space using a distributed representation, this paper hypothesizes that each sense of a word can be represented by one or more specific dimensions, and thus the target of aspect-level sentiment classification can be simplified into searching the related dimensions for the aspects and sentiments concerned. Particularly, an Attention Vector (ATV) based on attention mechanism is designed for each aspect in terms of a specific task, which involves two sub-vectors, i.e., a Dimension Attention Vector (DATV) and a Sentiment Attention Vector (SATV). The DATV determines the significances of different dimensions based on their correlations with an aspect; and the SATV allocates weights for the attributes of words, which are decided by sentiment polarities and part-of-speech (PoS) tagging. Given a sub-dataset related to an aspect, the ATV will be optimized by an Artificial Bee Colony (ABC) algorithm with a Support Vector Machine (SVM) classifier, the objective of which is to maximize classification accuracy. Intrinsically, the DATV can reduce the ambiguity existing in polysemy, meanwhile, the SATV is an auxiliary means for the optimization of the DATV, which can help eliminate the misunderstandings caused by antonyms. Then, the optimized DATV will be applied on a Convolutional Neural Network (CNN) model via simply scaling the pretrained word embeddings as inputs (named as ATV-CNN model). Experimental results show that the ATV-CNN model can have substantial advantages when compared with state-of-the-art models.

The contemporary unsupervised word representation methods have been successful in capturing semantic statistics on various Natural Language Processing tasks. However, these methods proved to be futile in addressing tasks like polysemy or homonymy, which prevail in such tasks. There has been a rise in the number of state-of-the-art transfer learning techniques bringing into play the language models pre-trained on large inclusive corpus. Motivated by these techniques, the present paper proposes an efficacious transfer learning based ensemble model. This model is inspired by ULMFit and presents results on challenging sentiment analysis tasks such as contextualization and regularization. We have empirically validated the efficiency of our proposed model by applying it to three conventional datasets for sentiment classification task. Our model accomplished the state-of-the-art outcomes remarkably when compared to acknowledged baselines in terms of classification accuracy.

The topic model has been widely applied to various research domains such as information retrieval, data mining, and so on. It can discover topics of texts in an unsupervised way. In the early years, most researches mainly focused on long texts. With the emergence of the Internet, the number of short texts is growing rapidly. Most existing schemes to solve the sparsity problems of short texts, are mainly based on data aggregation or model improvements. Among them, the Biterm Topic Model is one of the most representative models. It proposed a new way to model topics based on document-level word pairs and has shown creativity and effectiveness. However, this strategy ignores those semantically similar and rarely co-occurrent word pairs. What’s more, most researches ignore the multi-sense phenomenon in natural languages. In this paper, we utilize multi-sense word vectors to extract similar word pairs from the whole corpus by considering multiple senses. Based on this idea, we introduce a novel short-text topic model, which disambiguates multiple senses of words and generates more reasonable global biterms. Experimental results on two open-source English datasets have shown superiority to state-of-the-art topic models.

Determining the similarity between short texts plays an important role in natural language processing applications such as search, query suggestion and automatic summary, which has attracted widespread attention. Unlike traditional long texts, short texts present the characteristics of short length, weak signal, and high ambiguity. Researchers have proposed many methods, from simple vector space models to more sophisticated distributed semantics. However, these methods only consider the literal meaning of words, ignoring the impact of word ambiguity and the semantic information contained in the structure of the short text. Additionally, words on their own are often insufficient for expressing semantics, as many terms are composed of multiple words. In this paper, we propose a method based on semantic and syntactic information for short text similarity calculations by using knowledge and corpora to express the meaning of the term to solve polysemy, and using a constituency parse tree to capture the syntactic structure of short texts. Additionally, the proposed method uses terms as semantic units. Experimental results on ground-truth datasets demonstrate that the proposed method outperforms baseline methods.

Along with the emergence of the Internet, the rapid development of handheld devices has democratized content creation due to the extensive use of social media and has resulted in an explosion of short informal texts. Although a sentiment analysis of these texts is valuable for many reasons, this task is often perceived as a challenge given that these texts are often short, informal, noisy, and rich in language ambiguities, such as polysemy. Moreover, most of the existing sentiment analysis methods are based on clean data. In this paper, we present D I C E T , a transformer-based method for sentiment analysis that encodes representation from a transformer and applies deep intelligent contextual embedding to enhance the quality of tweets by removing noise while taking word sentiments, polysemy, syntax, and semantic knowledge into account. We also use the bidirectional long- and short-term memory network to determine the sentiment of a tweet. To validate the performance of the proposed framework, we perform extensive experiments on three benchmark datasets, and results show that D I C E T considerably outperforms the state of the art in sentiment classification.

Cross-lingual information retrieval (CLIR) allows user query in a different language from the language of target resources. Thus, translation is the key element in the query processing. There are three translation approaches: query, document, or hybrid query-document. However, query translation is very challenging due to the polysemy problem. Different linguistic nature of the languages will lead to ambiguity of meaning subsequently user’s true intention could be misinterpreted. This paper presents a semantic technique on query translation for a multilingual knowledge repository to improve the query processing. Offline translated documents or parallel corpora in English, Arabic, and Malay language including Jawi text was used as the data. Set of keywords were constructed preidentified by expert related to prophetic food. These keywords were annotated with the relevant Quranic verses, Hadith texts, Manuscript text images and scientific article determined by expert. The synonym and context-based translation was annotated together with the specific keyword. A query will do a three-way pattern match based on the keyword indexing list that link to the relevant documents. A one-stop knowledge repository on prophetic food was developed as a proof of concept using sources are from al-Quran, Hadith, classical manuscript, and scientific articles verified by experts to ensure the content authenticity and integrity.

In natural language, the phenomenon of polysemy is widespread, which makes it very difficult for machines to process natural language. Word sense disambiguation is a key issue in the field of natural language processing. This paper introduces the more common statistical learning methods used in the field of word sense disambiguation. Using the naive Bayesian machine learning method and the feature vector set extracted and constructed by the Dice coefficient method, a semantic word disambiguation model based on semantics is realized. The results of comparative experiments show that the proposed method is better compared with known systems. This paper proposes a method for disambiguation of word segmentation in professional fields based on unsupervised learning. This method does not rely on professional domain knowledge and training corpus and only uses the frequency, mutual information, and boundary entropy information of the string in the test corpus to solve the problem of word segmentation ambiguity. The experimental results show that these three evaluation standards can solve the problem of word segmentation ambiguity in professional fields and improve the effect of word segmentation. Among them, the segmentation result using mutual information is the best, and the performance is stable.

As a written language used for thousands of years, Ancient Chinese has some special characteristics like complex semantics as polysemy and the one-to-many alignment with Modern Chinese. Thus it may be translated in a large number of fully different but equally correct ways. In the absence of multiple references, reference-dependent evaluations like Bilingual Evaluation Understudy (BLEU) cannot identify potentially correct translation results. The explore on automatic evaluation of Ancient-Modern Chinese Translation is completely lacking. In this paper, we proposed an automatic evaluation metric for Ancient-Modern Chinese Translation called DTE (Dual-based Translation Evaluation), which can be used to evaluate one-to-many alignment in the absence of multiple references. When using DTE to evaluate, we found that the proper nouns often could not be correctly translated. Hence, we designed a new word segmentation method to improve the translation of proper nouns without increasing the size of the model vocabulary. Experiments show that DTE outperforms several general evaluations in terms of similarity to the evaluation of human experts. Meanwhile, the new word segmentation method promotes the Ancient-Modern Chinese translation models perform better on proper nouns’ translation, and get higher scores on both BLEU and DTE.

Polysemy, when a single term has multiple meanings, and synonymy, when multiple terms have the same meaning, are common phenomena in linguistics as well as in scientific knowledge. In ontology engineering, it is vital to detect the synonyms annotations and the multiple inheritances because of polysemy. The persistence of these issues in the semantic description of a knowledge domain causes problematic interoperability and data processing. The disambiguation of the entities, properties and relationships sense in a semantic web ontology significantly improves linked data generation and information retrieval. We explore the synonymy and polysemy in the setting of a cardiology terminology generated from textbooks on the basis of field coverage, professionals’ associations’ recommendations and bibliometrics, for the building of a cardiologic ontology. From 56,134 terms collected we found that 67.7% were unique. The indexed terms included single words, compound words and multi-word expressions. The frequency of their appearances in the combined master index was calculated and used as a marker of their significance. To cope with the linguistic polysemy and synonymy of terms, we examined them in WordNet, MeSH and BioPortal, as well as by latent semantic analysis (LSA) through singular value decomposition (SVD). Through these approaches we managed to identify and decipher semantic associations and relationships between the terms. We proposed a roadmap for ontology building from scratch by utilizing intrinsic and extrinsic knowledge resources and reuse of metadata. We anticipate that this approach is applicable in ontology engineering of different knowledge domains for relationships setting and linked data contextualization

The rapidly increasing volume of medical text data, including biomedical literature and clinical records, presents difficulties to biomedical researchers and clinical practitioners. Automatic text classification is an important means for managing medical text data. The main challenge in medical text classification is the complex terminology used in these documents. Therefore, it is critical to handle synonymy, polysemy, and multi-word concepts so that classification is based on the meaning of these documents. The solution to this problem of complex terminology helps in building systems with better access to relevant data, resulting in more effective utilisation of the existing information. In this paper, we present a simple and effective approach to address this challenge. A concept graph is automatically constructed and enriched for each medical text document with the help of a domain-specific similarity matrix that is built using Unified Medical Language System (UMLS) concepts in the training documents. Medical text documents are compared based on their enriched concept graphs using a graph kernel. Classification is then done based on the comparison result. The benefit of this approach is that it allows the incorporation of domain knowledge into the classification framework. The experiments on biomedical abstracts and clinical reports classification show the effectiveness of the proposed approach. Based on evaluation metrics of precision, recall and F1-scores, our method achieves a significantly higher classification performance than other widely used similarity measures for similarity-based text classification.

Well is an exemplary multi-functional word performing pragmatic and syntactic functions. That multi-functionality poses a potential problem: How do hearers in conversation determine which function is actualized and how do speakers project the function actualized? We address both questions examining factors hearers rely on to disambiguate well and the resources speakers deploy to designate well’s function. The study is based on 8-, 9-, and 10-word turns containing well extracted from the British National Corpus for which audio files from the Audio BNC are available. We include duration, measuring well’s durations in Praat. The workflow comprised both qualitative and quantitative methods. Qualitatively, all turns were manually inspected and the functions and subfunctions of well were identified. Due to data paucity the quantitative analysis was based only on a broad distinction between syntactic and pragmatic functions. The analysis involved two logistic regression model selection processes, one adopting a hearer, one a speaker perspective. Based on the factors position in the turn, duration and lexical context, our final models indicate that hearers disambiguate the two main functions of well drawing on lexical context and position in the turn while speakers project well’s functions by modulating duration. We propose that Hoey’s (2005) 6th priming hypothesis, concerned with polysemy, can be extended to also include polyfunctionality. Position also suggests a reading in terms of Hoey’s ‘textual colligation’ hypothesis related to a word’s position: particularly in its incarnation as a marker of dispreferreds, pragmatic well is heavily primed to occur turn-initially.

This paper aims to investigate the effectiveness of teaching polysemous nouns through an innovative method that the researcher calls Similarity-Based Approach (SBA) and to compare it to the translation-based vocabulary instruction method (TBM). The objective of the study is to show how guiding students’ comprehension of a polysemous word through semantic relations improves their understanding. At the same time, it is the aim of the study to demonstrate the effectiveness of Similarity-Based instruction by teaching vocabulary through showing semantic relation of word meaning to young learners compared to the old, traditional way of memorizing pairs of words. The experiment is conducted with two groups of elementary students with a pre-test, a lesson and a post-test. The results in the post-test revealed that through SBA students remember the vocabulary better than they do through the traditional method of memorization. SBA might represent a new method in teaching polysemous vocabulary to young learners, while TBM implies a traditional, well rooted way of teaching polysemous words in ESL classrooms.

To objectively and quantitatively study transcribed protocols of design conversations, we apply a semantic analysis approach based on dynamic semantic networks of nouns. We examined the applicability of the approach focused on a dynamic evaluation of the design problem solving process in engineering design educational settings. Using a case of real-world case, we show that the approach is able to determine the time dynamics of semantic factors such as level of abstraction, polysemy, information content, and quantify convergence/divergence in engineering design conversations.

This paper reports on a quasi-experimental study investigating the effects of a teaching methodology that combines Cognitive Linguistics (CL) and Corpus Linguistics (CrpL) insights on the acquisition of the two sets of English spatial prepositions of verticality, over/above and under/below. In addition, it is also concerned with learners’ attitudes towards the method of instruction. A total of 55 Moroccan first-year university students participated in this study, 22 of which were assigned to the experimental group, and 26 to the control group. The experimental group received what we termed the IPDDL instruction, which involves cognitive representations of the prepositions based on Image-schemas, the Principled Polysemy model, and Data-Driven Learning. The control group, in contrast, received a traditional instruction based on definitions and example sentences provided in dictionaries. Participants’ performance was measured with pre-, post-, and delayed post-tests in the form of a semantic judgment task and a gap-filling task. The findings demonstrate that the IPDDL method of instruction was significantly more effective in helping students acquire vertical prepositions, and its advantages proved durable posterior to two weeks of the instruction period. Thus, this study lends support to the applicability of a CL-CrpL-inspired methodology in second language instruction.

In the history of natural language processing (NLP) development, the representation of words has always been a significant research topic. In this survey, we provide a comprehensive typology of word representation models from a novel perspective that the development from static to dynamic embeddings can effectively address the polysemy problem, which has been a great challenge in this field. Then the survey covers the main evaluation metrics and applications of these word embeddings. And, we further discuss the development of word embeddings from static to dynamic in cross-lingual scenario. Finally, we point out some open issues and future works.

In this paper, we present a formalism for handling polysemy in spatial expressions based on supervaluation semantics called standpoint semantics for polysemy (SSP). The goal of this formalism is, given a prepositional phrase, to define its possible spatial interpretations. For this, we propose to characterize spatial prepositions by means of a triplet $\langle $image schema, semantic feature, spatial axis$\rangle $. The core of SSP is predicate grounding theories, which are formulas of a first-order language that define a spatial preposition through the semantic features of its trajector and landmark. Precisifications are also established, which are a set of formulae of a qualitative spatial reasoning formalism that aims to provide the spatial characterization of the trajector with respect to the landmark. In addition to the theoretical model, we also present results of a computational implementation of SSP for the preposition ‘in’.

Event extraction is a challenging problem in information extraction, designed to extract structured information from unstructured text. The existing event extraction methods are mostly based on the pipeline model and use the traditional language representation model for word embedding. It is impossible to model the polysemy and the features is very simple, besides, the pipeline model is also easy to generate cascading errors. Aiming at the above problems, we propose a joint event extraction model, which combines part-of-speech features, entity features and word representation of BERT model. Because of the BERT can’t capture local features, we use multi-CNN to capture the local features of different scales. and further use BiLSTM to capture context features to assist joint event extraction. Experimental results show that the precision, recall and F1 of the method on the CEC dataset are 88.83%, 90.91% and 89.86%, respectively, which are 1.61%, 3.6% and 2.6% higher than the PLMEE model.

Most current automatic summarization methods are for English texts. The distinction between words in Chinese text is large, the types of parts of speech are many and complex, and polysemy or ambiguous words appear frequently. Therefore, compared with English text, Chinese text is more difficult to extract useful feature words. Due to the complex syntax of Chinese, there are currently relatively few automatic summarization methods for Chinese text. In the past, only the important sentences in the original text can be selected and simply arranged to obtain a summary with chaotic sentences and insufficient coherence. Meanwhile, because Chinese short text usually contains more redundant information and the sentence structure is not neat, we propose a topic-based automatic summary method for Chinese short text. Firstly, a key sentence selection method is proposed combining topic words and TF-IDF to obtain the score of each text corresponding to the topic in the original text data. Then the sentence with the highest score as the topic sentence of the topic is selected. Considering that the short text of Weibo may contain a lot of irrelevant information and sometimes even lack some important components of topic, three retouching mechanisms are proposed to improve the conciseness, richness and readability of topic sentence extraction results. We validate our approach on natural disaster and social hot event datasets from Sina Weibo. The experimental results show that the polished topic summary not only reflects the exact relationship between topic sentences and natural disasters or social hot events, but also has rich semantic information. More importantly, we can almost grasp the basic elements of natural disaster or social hot event from the topic sentence, so as to help the government guide disaster relief or meet the needs of users for quickly obtaining information of social hot events.

The task of relation recognition identifies semantic relationships between two named entities in a sentence. In neural network based models, a convolutional layer is often conducted to extract representative local features of a sentence. The convolution operation is implemented through a whole sentence, without considering the structure of a sentence. Because the task to recognize entity relation is processed in sentence level, many ambiguous phenomena (e.g., polysemy) are influential rather than in a document. Capturing structural information of a sentence is helpful to solve this problem. In this paper, a multi-channel framework is presented, which uses two named entities to divide a sentence into several channels. Each channel is stacked with layered neural networks. These channels do not interact during recurrent propagation, which enables a neural network to learn different representations. In our experiments, it outperforms the widely used position embedding approach. Comparing with the state-of-the-art approaches, its performance shows a meaningful improvement.

A pun is always humorous and has strong interactive value in people's daily communication. It creates a humorous effect in a certain context, in which a word implies two or more meanings by using polysemy (homographic pun) or phonological similarity to another word (heterographic pun). Pun location is a task to identify the pun word in a given text, which is of great significance to understand humorous texts. Existing methods generally adopt single long sequence structure but cannot well capture the rich semantics of pun words in sentences. We present an approach that considers long-distance and short-distance semantic relations between words simultaneously. For the long-distance semantic relation, we introduce multi-level embeddings to represent the most relevant aspects of the data. For the short-distance semantic relation, we exploit the complex-valued model with a self-adaptive selection mechanism based on multi-scale of input information. Meanwhile, we propose a new classification task to distinguish the homographic pun and heterographic pun. We introduce it as an auxiliary to jointly train the original pun location task, which first learns the location of different types of puns together. Experiment results show that the latest state-of-the-art results can be achieved through our model.

In Natural Language Processing, determining the semantic likeness between sentences is an important research area. For example, there exists many possible semantics for a word (polysemy), and the synonym of the word differs. Double LSTM (Long Short Term Memory) working at same time on double phrase sequences model is projected to overcome the solitary sequence problem. Furthermore, with the goal of overcoming the second issue, as indicated by the qualities of English dialect, we utilized the British corpus semantic similarity datasets structured by specialists to prepare, and validate the technique. During the training process the stopwords were reserved for use. Convolution Neural Network and Semantic Likeness model based on grammar are used to compare the results of our projected representation. The outcomes demonstrate that the proposed methodology is more prominent than the previous approaches by means of precision, recall rate, accuracy etc., along with the enhanced generalization potential of the neural network.

Human creativity generates novel ideas to solve real-world problems. This thereby grants us the power to transform the surrounding world and extend our human attributes beyond what is currently possible. Creative ideas are not just new and unexpected, but are also successful in providing solutions that are useful, efficient and valuable. Thus, creativity optimizes the use of available resources and increases wealth. The origin of human creativity, however, is poorly understood, and semantic measures that could predict the success of generated ideas are currently unknown. Here, we analyze a dataset of design problem-solving conversations in real-world settings by using 49 semantic measures based on WordNet 3.1 and demonstrate that a divergence of semantic similarity, an increased information content, and a decreased polysemy predict the success of generated ideas. The first feedback from clients also enhances information content and leads to a divergence of successful ideas in creative problem solving. These results advance cognitive science by identifying real-world processes in human problem solving that are relevant to the success of produced solutions and provide tools for real-time monitoring of problem solving, student training and skill acquisition. A selected subset of information content (IC S\'anchez-Batet) and semantic similarity (Lin/S\'anchez-Batet) measures, which are both statistically powerful and computationally fast, could support the development of technologies for computer-assisted enhancements of human creativity or for the implementation of creativity in machines endowed with general artificial intelligence., Comment: 42 pages, 10 figures, 2 tables

For the Chinese text sentiment classification task, the preprocessing based on deep learning models cannot retain the information and polysemy of the word in the sentence well. So this paper adopts the newly developed ERNIE [1–2] (Knowledge Enhanced Semantic Representation) pre-training model from Baidu, which is based on word feature input modeling, not only enhances the semantic representation of the word, but also preserves the contextual information of the word and the polysemy of the word. After pre-training by ERNIE model, the output word vector is used as the input of BiLSTM (bidirectional long and short-term memory network) model for training and obtaining sentiment classification results. The accuracy rate of Ernie bilstm model is 92.35% after verification on nlpcc2014 microblog sentiment analysis sample data set, which proves that the model has good performance in Chinese text sentiment classification task.

The topic discovery method, as an effective tool for semantic mining and a key means to extract new features from original text, plays an important role in the field of text mining and knowledge discovery. To solve the problems encountered in traditional topic models, such as the loss of semantic information and the ambiguity of topic concepts, as well as the crossover and coverage among topics, we propose a semantic topic discovery method based on the conditional co-occurrence degree (CCOD_STDM). First, every document is split into multiple subdocuments according to the semantic structure of the document and the independence decision rules. Second, combinatorial words with strong semantic relevance are extracted based on the conditional co-occurrence degree within the subdocuments. Based on these combinatorial words, new subdocuments are formed by feature expansion and content reconstruction. Third, “topic-word” distributions and “document-topic” distributions of new subdocuments are obtained by topic modeling with Gibbs sampling. Finally, “document-topic” distributions of the original documents are obtained by merging new subdocuments’ “document-topic” distributions with specific strategies. The numerical experiments are compared with six topic models and two evaluation methods on seven kinds of public corpora, and the experimental results verify the superiority of CCOD_STDM and its efficiency in topic discovery. More importantly, a case study illustrates that the combinatorial words can effectively avoid the polysemy problem and can facilitate the condensation and summary of topics.

Recent years have seen the success of applying word embedding algorithms to natural language processing (NLP) tasks. Most word embedding algorithms only produce a single embedding per word. This makes the learned embeddings indiscriminative since many words are polysemous. Some prior work utilizes the context in which the word resides to learn multiple word embeddings. However, context-based solutions are problematic for short texts, such as tweets, which have limited context. Moreover, existing approaches tend to enumerate all possible context types of a particular word regardless of their target applications. Applying multiple vector representations per word in NLP tasks can be computationally expensive because all possible combinations of senses of words in a snippet need to be considered. Sometimes, a word sense can be captured when the class information or label of the short text is presented. For example, in a disaster-related dataset, when a text snippet is labeled as “hurricane related”, the word “water” in the snippet is more likely to be interpreted as rain and flood; when a snippet is labeled as “hurricane unrelated”, the word “water” can be interpreted as its more general meaning. In this work, we propose to use class information to enhance the discriminativeness of words. Instead of enumerating all potential senses per word in the text, the number of vector representations per word should be a function of the future classification task. We show that learning the number of vector representations per word according to the number of classes in the classification task is often sufficient to clarify the polysemy. Word embeddings learned from neural language models typically have the property of good linear compositionality. We utilize this property to encode class information into the vector representation of a word. We explore four approaches to train class-specific embeddings to encode class information by utilizing the label information and the linear compositionality property of word embeddings. We present a general framework consisting of a pair of convolutional neural networks to utilize the learned class-specific word embeddings as input for text classification tasks. We evaluate our approach and framework on topic classification of a disaster-focused Twitter dataset and a benchmark Twitter sentiment classification dataset from SemEval 2013. Our results show a relative accuracy improvement of 3–4% over a recent baseline.

Natural language processing (NLP) is one of the key techniques in intelligent question-answering (Q&A) systems. Although recurrent neural networks and long short-term memory (LSTM) networks exhibit obvious advantages on well-known English Q&A datasets, they still suffer from several defects including indeterminateness, polysemy and the lack of changing morphology in Chinese, which results in complex NLP on large and diverse Chinese Q&A datasets. In this paper, we first analyze limitations of applying LSTM and bidirectional LSTM (Bi-LSTM) models to noisy Chinese Q&A datasets. Then, we focus on integrating attention mechanisms and multi-granularity word segmentation into Bi-LSTM and propose an attention mechanism and multi-granularity-based Bi-LSTM model (AM–Bi-LSTM) which combines the improved attention mechanism with a novel processing of multi-granularity word segmentation to handle the complex NLP in Chinese Q&A datasets. Furthermore, similarity of questions and answers is formulated to implement the quantitative computation which helps to achieve better performance in Chinese Q&A systems. Finally, we verify the proposed model on a noisy Chinese Q&A dataset. The experimental results demonstrate that the novel AM–Bi-LSTM model achieves significant improvement on evaluation metrics of accuracy, mean average precision and so on. Moreover, the experimental results indicate that the novel AM–Bi-LSTM model outperforms baseline methods and other LSTM-based models.

This paper offers a step-by-step guide to compiling a dictionary entry by applying cognitive semantic theories to corpus data. In this way, the paper aims to make a contribution to the field of cognitive lexicography. The focus of attention is on the lexicographic tasks of identifying lexical units, capturing usage patterns and providing definitions. In an attempt to systematize these demanding and highly subjective tasks, we combine the principles of corpus linguistics (corpus pattern analysis) and cognitive semantics (frame semantics, conceptual metaphor and metonymy theory, principled polysemy approach). The proposed integrated (corpus-based and cognitively oriented) lexicographic approach is demonstrated in the analysis of a polysemous lexical item from the semantic field of motion, the verb float. The independent semantic and phraseological analysis of float is compared with the corresponding entries of monolingual learners’ dictionaries to specify in what respects the described approach can improve EFL lexicography. The paper is not restricted to a theoretical discussion of lexicographic issues or a critical review of existing entries; rather, a new version of the float entry is offered. The proposed entry is claimed to be more transparent and streamlined than existing ones and can serve learners better in both decoding and encoding tasks.

To objectively and quantitatively study transcribed protocols of design problem solving conversations, we propose a semantic analysis approach based on dynamic semantic networks of nouns constructed with WordNet 3.1 lexical database. We examined the applicability of the semantic approach focused on a dynamic evaluation of the design problem solving process in educational settings. Using a case of real- world design problem-solving conversations, we show that the approach is able to determine the time dynamics of semantic factors such as level of abstraction, polysemy or information content, and quantify convergence/divergence of semantic similarity in design conversations between students, instructors and real clients. The approach can also be used to evaluate the aforementioned semantic factors for successful and unsuccessful ideas generated in the process of design problem solving, or to assess the effect of external feedback on the developed design solution. The proposed semantic analysis approach allows fast computation of the semantic factors in real time thereby demavonstrating a potential for both monitoring and support of the design problem solving process.