Word Segmentation Research Articles

Clear speech improves word segmentation in quiet and in noise: Evidence from visual-world eye-tracking

Listener-oriented hyperarticulated clear speech facilitates linguistic processing and cognitive functioning associated with speech perception under various listening conditions. Using the visual-world eye-tracking paradigm, we investigated whether clear speech also aids speech segmentation, or the discovery of word boundaries, and examined the dynamic time course of its effect. Native American English speakers (N = 77) heard sentences in which the target word (e.g., ham) was temporarily ambiguous with a longer unintended competitor (e.g., hamster) across a word boundary (e.g., She saw the ham starting…) while viewing images depicting the target, competitor, and unrelated distractors. Clear and conversational sentences were presented in quiet or in speech-shaped noise at +3 dB signal-to-noise ratio. Analysis of eye fixations to the images over time revealed that compared with conversational speech, clear speech facilitated the disambiguation of the target from the competitor even before the disambiguation point was reached. The facilitation was found in both listening conditions but was relatively delayed in noise. These findings suggest that speaking clearly improves word segmentation and reduces lexical competition especially in optimal listening conditions. The speech segmentation facilitation may partly underlie the clear speech benefits observed for other signal-dependent and relatively signal-independent linguistic and cognitive processes.

Big Data Analysis and Forecast of Employment Position Requirements for College Students

With the help of natural language processing and machine learning, we can analyze the information of online recruitment text posted by employer companies and dig the requirement features of these employment positions, and this is a meaningful work with practical value. However, existing methods for analyzing online recruitment information are too simple to withstand the mass data on the Internet, so this paper aims to study the analysis and forecast of employment position requirements for college students based on big data analysis. At first, the recruitment information of companies is preprocessed, keywords in the recruitment text are extracted by the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm in the Python Chinese word segmentation toolkit, words in the recruitment text are segmented using the open source tool Word2Vec, and the uncounted words in the recruitment text are identified based on the Conditional Random Field (CRF) model. Then, this paper compares the occurrence probability of skills learnt by college students in a certain company employment position with the occurrence probability of the skills in all employment positions on the website, so as to find out the core skills learnt by college students that can match with the job positions required by companies. At last, this paper builds a XGBoost model to forecast the employment position requirements for college students, and verifies the validity of the model using experimental results.

A Multi-Granularity Word Fusion Method for Chinese NER

Named entity recognition (NER) plays a crucial role in many downstream natural language processing (NLP) tasks. It is challenging for Chinese NER because of certain features of Chinese. Recently, large-scaled pre-training language models have been used in Chinese NER. However, since some of the pre-training language models do not use word information or just employ word information of single granularity, the semantic information in sentences could not be fully captured, which affects these models’ performance. To fully take advantage of word information and obtain richer semantic information, we propose a multi-granularity word fusion method for Chinese NER. We introduce multi-granularity word information into our model. To make full use of the information, we classify the information into three kinds: strong information, moderate information, and weak information. These kinds of information are encoded by encoders and then integrated with each other through the strong-weak feedback attention mechanism. Specifically, we apply two separate attention networks to word embeddings and N-grams embeddings. Then, the outputs are fused into another attention. In these three attentions, character embeddings are used to be the query of attentions. We call the results the multi-granularity word information. To combine character information and multi-granularity word information, we introduce two fusion strategies for better performance. The process makes our model obtain rich semantic information and reduces word segmentation errors and noise in an explicit way. We design experiments to get our model’s best performance by comparing some components. Ablation study is used to verify the effectiveness of each module. The final experiments are conducted on four Chinese NER benchmark datasets and the F1 scores are 81.51% for Ontonotes4.0, 95.47% for MSRA, 95.87% for Resume, and 69.41% for Weibo. The best improvement achieved by the proposed method is 1.37%. Experimental results show that our method outperforms most baselines and achieves the state-of-the-art method in performance.

Exploring the Potential of BERT-BiLSTM-CRF and the Attention Mechanism in Building a Tourism Knowledge Graph

As an important infrastructure in the era of big data, the knowledge graph can integrate and manage data resources. Therefore, the construction of tourism knowledge graphs with wide coverage and of high quality in terms of information from the perspective of tourists’ needs is an effective solution to the problem of information clutter in the tourism field. This paper first analyzes the current state of domestic and international research on constructing tourism knowledge graphs and highlights the problems associated with constructing knowledge graphs, which are that they are time-consuming, laborious and have a single function. In order to make up for these shortcomings, this paper proposes a set of systematic methods to build a tourism knowledge graph. This method integrates the BiLSTM and BERT models and combines these with the attention mechanism. The steps of this methods are as follows: First, data preprocessing is carried out by word segmentation and removing stop words; second, after extracting the features and vectorization of the words, the cosine similarity method is used to classify the tourism text, with the text classification based on naive Bayes being compared through experiments; third, the popular tourism words are obtained through the popularity analysis model. This paper proposes two models to obtain popular words: One is a multi-dimensional tourism product popularity analysis model based on principal component analysis; the other is a popularity analysis model based on emotion analysis; fourth, this paper uses the BiLSTM-CRF model to identify entities and the cosine similarity method to predict the relationship between entities so as to extract high-quality tourism knowledge triplets. In order to improve the effect of entity recognition, this paper proposes entity recognition based on the BiLSTM-LPT and BiLSTM-Hanlp models. The experimental results show that the model can effectively improve the efficiency of entity recognition; finally, a high-quality tourism knowledge was imported into the Neo4j graphic database to build a tourism knowledge graph.

Consequences of phonological variation for algorithmic word segmentation

Over the first year, infants begin to learn the words of their language. Previous work suggests that certain statistical regularities in speech could help infants segment the speech stream into words, thereby forming a proto-lexicon that could support learning of the eventual vocabulary. However, computational models of word segmentation have typically been tested using language input that is much less variable than actual speech is. We show that using actual, transcribed pronunciations rather than dictionary pronunciations of the same speech leads to worse segmentation performance across models. We also find that phonologically variable input poses serious problems for lexicon building, because even correctly segmented word forms exhibit a complex, many-to-many relationship with speakers' intended words. Many phonologically distinct word forms were actually the same intended word, and many identical transcriptions came from different intended words. The fact that previous models appear to have substantially overestimated the utility of simple statistical heuristics suggests a need to consider the formation of the lexicon in infancy differently.

CLASSIC Utterance Boundary: A Chunking‐Based Model of Early Naturalistic Word Segmentation

AbstractWord segmentation is a crucial step in children's vocabulary learning. While computational models of word segmentation can capture infants’ performance in small‐scale artificial tasks, the examination of early word segmentation in naturalistic settings has been limited by the lack of measures that can relate models’ performance to developmental data. Here, we extended CLASSIC (Chunking Lexical and Sublexical Sequences in Children; Jones et al., 2021), a corpus‐trained chunking model that can simulate several memory and phonological and vocabulary learning phenomena to allow it to perform word segmentation using utterance boundary information, and we have named this extended version CLASSIC utterance boundary (CLASSIC‐UB). Further, we compared our model to the performance of children on a wide range of new measures, capitalizing on the link between word segmentation and vocabulary learning abilities. We showed that the combination of chunking and utterance‐boundary information used by CLASSIC utterance boundary allowed a better prediction of English‐learning children's output vocabulary than did other models.

A Levenshtein distance-based method for word segmentation in corpus augmentation of geoscience texts

ABSTRACT For geoscience text, rich domain corpora have become the basis of improving the model performance in word segmentation. However, the lack of domain-specific corpus with annotation labelled has become a major obstacle to professional information mining in geoscience fields. In this paper, we propose a corpus augmentation method based on Levenshtein distance. According to the technique, a geoscience dictionary of 20,137 words was collected and constructed by crawling the keywords from published papers in China National Knowledge Infrastructure (CNKI). The dictionary was further used as the main source of synonyms to enrich the geoscience corpus according to the Levenshtein distance between words. Finally, a Chinese word segmentation model combining the BERT, Bi-gated recurrent neural network (Bi-GRU), and conditional random fields (CRF) was implemented. Geoscience corpus composed of complex long specific vocabularies has been selected to test the proposed word segmentation framework. CNN-LSTM, Bi-LSTM-CRF, and Bi-GRU-CRF models were all selected to evaluate the effects of Levenshtein data augmentation technique. Experiments results prove that the proposed methods achieve a significant performance improvement of more than 10%. It has great potential for natural languages processing tasks like named entity recognition and relation extraction.

The impact of explicitly teaching word segmentation using a visual vowel hand sign system to help at-risk children learn to read and spell English: A proof-of-concept study

ABSTRACT English vowels are phonologically and orthographically more difficult than consonants when learning to map speech to print. We sought to determine if teaching young at-risk readers and spellers to use a visual vowel hand sign system to segment spoken words into their component phonemes contributed to improved grapheme-phoneme correspondence knowledge, word reading and spelling. Twenty-three at-risk Year 1 participants were assessed prior to receiving Crack the Code – an intervention that incorporates a visual vowel hand sign system for phoneme segmentation – and immediately after its completion on measures of segmentation, grapheme-phoneme correspondence knowledge, word reading accuracy, word reading efficiency and spelling. Analysis of group data showed statistically significant improvement across all measures with effect sizes ranging from 0.73 to 1.83. Explicitly teaching phonemic segmentation with a visual vowel sign system may be an effective way of improving vowel knowledge, word reading and spelling skills of at-risk children.

Word Segmentation on Discovered Phone Units With Dynamic Programming and Self-Supervised Scoring

Recent work on unsupervised speech segmentation has used self-supervised models with phone and word segmentation modules that are trained jointly. This paper instead revisits an older approach to word segmentation: bottom-up phone-like unit discovery is performed first, and symbolic word segmentation is then performed on top of the discovered units (without influencing the lower level). To do this, I propose a new unit discovery model, a new symbolic word segmentation model, and then chain the two models to segment speech. Both models use dynamic programming to minimize segment costs from a self-supervised network with an additional duration penalty that encourages longer units. Concretely, for acoustic unit discovery, duration-penalized dynamic programming (DPDP) is used with a contrastive predictive coding model as the scoring network. For word segmentation, DPDP is applied with an autoencoding recurrent neural as the scoring network. The two models are chained in order to segment speech. This approach gives comparable word segmentation results to state-of-the-art joint self-supervised segmentation models on an English benchmark. On French, Mandarin, German and Wolof data, it outperforms previous systems on the ZeroSpeech benchmarks. Analysis shows that the chained DPDP system segments shorter filler words well, but longer words might require some external top-down signal.

The importance of the positional probability of word final (but not word initial) characters for word segmentation and identification in children and adults' natural Chinese reading.

Word spacing is important in guiding eye movements during spaced alphabetic reading. Chinese is unspaced and it remains unclear as to how Chinese readers segment and identify words in reading. We conducted two parallel experiments to investigate whether the positional probabilities of the initial and the final characters of a multicharacter word affected word segmentation and identification in Chinese reading. Two-character words were selected as targets. In Experiment 1, the initial character's positional probability was manipulated as being either high or low, and the final character was kept identical across the two conditions. In Experiment 2, an analogous manipulation was made for the final character of the target word. We recorded adults' and children's eye movements when they read sentences containing these words. In Experiment 1, reading times on targets did not differ in the two conditions for both children and adults, providing no evidence that a word initial character's positional probability contributes to word segmentation. In Experiment 2, adults had shorter reading times and made fewer refixations on targets that comprised final characters with high relative to low positional probabilities; a similar effect was observed in children, but this effect had a slower time course. The results demonstrate that the positional probability of the final (but not the initial) character of a word influences segmentation commitments in reading. It suggests that Chinese readers identify where a currently fixated word ends, and via this commitment, by default, they identify where the subsequent word begins. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

LATTE: Lattice ATTentive Encoding for Character-based Word Segmentation

A character sequence comprises at least one or more segmentation alternatives. This can be considered segmentation ambiguity and may weaken segmentation performance in word segmentation. Proper handling of such ambiguity lessens ambiguous decisions on word boundaries. Previous works have achieved remarkable segmentation performance and alleviated the ambiguity problem by incorporating the lattice, owing to its ability to capture segmentation alternatives, along with graph-based and pre-trained models. However, multiple granularity information, including character and word, in a lattice that encodes with such models may not be attentively exploited. To strengthen multi-granularity representations in a lattice, we propose the Lattice ATTentive Encoding (LATTE) method for character-based word segmentation. Our model employs the lattice structure to handle segmentation alternatives and utilizes graph neural networks along with an attention mechanism to attentively extract multi-granularity representation from the lattice for complementing character representations. Our experimental results demonstrated improvements in segmentation performance on the BCCWJ, CTB6, and BEST2010 datasets in three languages, particularly Japanese, Chinese, and Thai.

Thai Word Segmentation with a Brain-Inspired Sparse Distributed Representations Learning Memory.

Word segmentation is necessary for many natural language processing, especially Thai language, that is, unsegmented words. However, wrong segmentation causes terrible performance in the final result. In this study, we propose two new brain-inspired methods based on Hawkins' approach to address Thai word segmentation. Sparse Distributed Representations (SDRs) are used to model the neocortex structure of the brain to store and transfer information. The first proposed method, THDICTSDR, improves the dictionary-based approach by utilizing SDRs to learn the surrounding context and combine with n-gram to select the correct word. The second method uses SDRs instead of a dictionary and is called THSDR. The evaluation uses the BEST2010 and LST20 standard datasets for segmentation words by comparing them with the longest matching, newmm, and Deepcut, which is state-of-the-art in the deep learning approach. The result shows that the first method provides the accuracy, and performances are significantly better than other dictionary bases. The first new method can achieve F1-Score at 95.60%, comparable to the state-of-the-art and Deepcut F1-Score at 96.34%. However, it provides a better performance F1-Score at 96.78% in learning all vocabularies. In addition, it can achieve 99.48% F1-Score beyond Deepcut 97.65% in case of all sentences being learnt. The second method has fault tolerance to noise and provides overall result over deep learning in all cases.

Character-Based Thai Word Segmentation with Multiple Attentions

Character-based word segmentation models have been extensively applied to Asian languages, including Thai, owing to their promising performance. These models estimate the word boundaries from a character sequence; however, a Thai character unit in a sequence has no inherent meaning, in contrast with word, subword, and character cluster units that represent more meaningful linguistic information. In this paper, we propose a Thai word segmentation model that uses various types of information, including words, subwords, and character clusters, from a character sequence. Our model applies multiple attentions to refine segmentation inferences by estimating the significant relationships among characters and various unit types. We evaluated our model on three Thai datasets, and the experimental results show that our model outperforms other Thai word segmentation models, demonstrating the validity of using character clusters over subword units. A case study on sample Thai text supported these results. Thus, according to our analysis, particularly the case study, our model can segment Thai text accurately, while other existing models yield incorrect results that violate the Thai writing system.

Peculiarities of the Arabic Language Processing: Morphological Modeling

The paper deals with the features of morphological modeling of the Arabic language based on the definition of the specifics of its formalization. Morphological modeling is one of the key stages of automatic text analysis and includes tools for building a word form to a stem, root, definition of a part of speech, automatic construction (generation) of a given word form, etc. The objectives of the study are interdisciplinary in nature and include both the theoretical aspects of studying the features of the Arabic language, which are most relevant for its automatic processing, and the study of existing morphological analyzers and determining the specifics of their work. The practical part is based on testing the CAMeL TOOLS, one of the advantages of which is its comprehensive nature, which allows both preprocessing of text and solving applied problems, including sentiment analysis. The criteria for selecting examples for testing took into account the features of the Arabic language, which are difficult for its formalization (segmentation of functional words with continuous spelling, morphological and lexical homonymy, etc.). The variability of the generalized concept of "the Arabic language" is taken into account, which combines classical Arabic, Modern Standard Arabic and modern Arabic dialects. Testing tools for morphological modeling allows us to draw conclusions about the need to improve the terminological apparatus, the variability of which is noted in the description of word forms. Such kind of variation (divergence from the concepts accepted in general linguistics) potentially leads to a distortion of the results of lexico-semantic analysis. During the analysis, some gaps were noted related to the definition of part-of-speech belonging, the description of word forms, etc. The results of the study are relevant both for linguistic research and for improving the development of software applications aimed at processing the Arabic text.

Visual News Ticker Surveillance Approach from Arabic Broadcast Streams

The news ticker is a common feature of many different news networks that display headlines and other information. News ticker recognition applications are highly valuable in e-business and news surveillance for media regulatory authorities. In this paper, we focus on the automatic Arabic Ticker Recognition system for the Al-Ekhbariya news channel. The primary emphasis of this research is on ticker recognition methods and storage schemes. To that end, the research is aimed at character-wise explicit segmentation using a semantic segmentation technique and words identification method. The proposed learning architecture considers the grouping of homogeneous-shaped classes. This incorporates linguistic taxonomy in a unified manner to address the imbalance in data distribution which leads to individual biases. Furthermore, experiments with a novel Arabic News Ticker (Al-ENT) dataset that provides accurate character-level and character components-level labeling to evaluate the effectiveness of the suggested approach. The proposed method attains 96.5%, outperforming the current state-of-the-art technique by 8.5%. The study reveals that our strategy improves the performance of low-representation correlated character classes.

Rotation, Translation and Scale Invariant Sign Word Recognition Using Deep Learning

Communication between people with disabilities and people who do not understand sign language is a growing social need and can be a tedious task. One of the main functions of sign language is to communicate with each other through hand gestures. Recognition of hand gestures has become an important challenge for the recognition of sign language. There are many existing models that can produce a good accuracy, but if the model test with rotated or translated images, they may face some difficulties to make good performance accuracy. To resolve these challenges of hand gesture recognition, we proposed a Rotation, Translation and Scale-invariant sign word recognition system using a convolutional neural network (CNN). We have followed three steps in our work: rotated, translated and scaled (RTS) version dataset generation, gesture segmentation, and sign word classification. Firstly, we have enlarged a benchmark dataset of 20 sign words by making different amounts of Rotation, Translation and Scale of the original images to create the RTS version dataset. Then we have applied the gesture segmentation technique. The segmentation consists of three levels, i) Otsu Thresholding with YCbCr, ii) Morphological analysis: dilation through opening morphology and iii) Watershed algorithm. Finally, our designed CNN model has been trained to classify the hand gesture as well as the sign word. Our model has been evaluated using the twenty sign word dataset, five sign word dataset and the RTS version of these datasets. We achieved 99.30% accuracy from the twenty sign word dataset evaluation, 99.10% accuracy from the RTS version of the twenty sign word evolution, 100% accuracy from the five sign word dataset evaluation, and 98.00% accuracy from the RTS version five sign word dataset evolution. Furthermore, the influence of our model exists in competitive results with state-of-the-art methods in sign word recognition.

Особенности морфемной структуры слов со связанным корнем

This article is devoted to the problem of segmentation of words with a bound radix. The subject of study is the morphemic articulation of bilaterally bound radix of Russian origin. The purpose of the article is to identify the reasons for the formation of connectedness on the example of historically cognate words with the radix *stig. The first part of the article contains a brief analysis of the main synchronic and diachronic studies in the field of bound radix. Groups of related radix and their features are revealed, the question of studying verbs with such radix as part of complex lexical systems — etymological nests and genetic paradigms is considered. The second part of the article is devoted to identifying the reasons for the formation of the connection of words with the radix *stig. As a result of the study, the archaization of the originator was revealed, which led to the rupture of semantic links between derivatives and led to the collapse of the etymological family of words.

Реликтовые корневые морфемы казахского языка в ряду универсалий

Lexical remnants as one of the factors of the dynamic processes of word formation, are caused by the language existence. These phenomena are included in the category of exceptions in the system of languages and called «fossilized constructions», «quasi-morphs», «dead roots», etc. The new approach to the morphemic segmentation of words with lexical remnants is represented in the article. The term “relic root morphemes” (RRM) is suggested for lexical remnants. 113Н. Букетова, А. Аратаева, С. Туркенова, А. Амренова Turkic Studies Journal 4 (2023) 99-114The example of RRM separation based on synchronic word formation analysis is given in the article. According to F. de Saussure, the RRM is considered as a linguistic sign, which has its meaning both from the point of view of the current language state and the historical perspective. The analysis of words with RRM based on diachronic approach makes it possible to identify at what level of time factor of the diachronic depth the meaning accumulation takes place.The existence of the RRM in languages, including Kazakh, has a universal character. The relic root morpheme is a residually distinguishable root morpheme, the semantics of which is characterized by idiomaticity, expressed in the word-forming meaning, and confirmed by multiple semantics, finding a place in different historical periods in diachronic aspect.

Matematičke kompetencije kao preduslov za izvođenje pojedinih pretčitalačkih vežbi sa decom predškolskog uzrasta

The emphasis of the research presented in this paper is on pre-reading exercises with preschool-age children and the exploration of logical-mathematical competences, which are considered prerequisites for performing these exercises. The study involved examining children aged five to six, where they were given pre-reading and mathematical tasks. The objective was to uncover potential interdependencies between these competences in terms of how they are solved and their relationship with children's abilities in both areas. This investigation aims to inspire methodological approaches for integrating these domains. Data was collected from twenty children through individual testing and interviews. The results we obtained indicate, among other findings, that the following mathematical competencies are essential for promoting early literacy: a) understanding 1:1 correspondence during vote counting, b) comparing word lengths by using a non-standard measuring unit, c) knowledge of ordinal numbers and positional relationships when determining the position of a sound within a word, and d) comprehension of sets (understanding the relationship between the whole and its parts) during the transition from the iconic representation of a complete word to the indexical representation of a word segment (sound).

Applying Blockchain Technology in Network Public Opinion Risk Management System in Big Data Environment.

Network public opinion represents public social opinion to a certain extent and has an important impact on formulating national policies and judgment. Therefore, China and other countries attach great importance to the study of online public opinion. However, the current researches lack the combination of theory and practical cases and lack the intersection of social and natural sciences. This work aims to overcome the technical defects of traditional management systems, break through the difficulties and pain points of existing network public opinion risk management, and improve the efficiency of network public opinion risk management. Firstly, a network public opinion isolation strategy based on the infectious disease propagation model is proposed, and the optimal control theory is used to realize a functional control model to maximize social utility. Secondly, blockchain technology is used to build a network public opinion risk management system. The system is used to conduct a detailed study on identifying and perceiving online public opinion risk. Finally, a Chinese word segmentation scheme based on Long Short-Term Memory (LSTM) network model and a text emotion recognition scheme based on a convolutional neural network are proposed. Both schemes are validated on a typical corpus. The results show that when the system has a control strategy, the number of susceptible drops significantly. Two days after the public opinion is generated, the number of susceptible people decreased from 1,000 to 250; 3 days after the public opinion is generated, the number of susceptible people stabilized. 2 days after the public opinion is generated, the number of lurkers increased from 100 to 620; 3 days after the public opinion is generated, the number of lurkers stabilized. The data demonstrate that the designed isolation control strategy is effective. Changes in public opinion among infected people show that quarantine control strategies played a significant role in the early days of Corona Virus Disease 2019. The rate of change in the number of infections is more affected when quarantine controls are increased, especially in the days leading up to the outbreak. When the system adopts the optimal control strategy, the influence scope of public opinion becomes smaller, and the control becomes easier. When the dimension of the word vector of emergent events is 200, its accuracy may be higher. This method provides certain ideas for blockchain and deep learning technology in network public opinion control.