Search results for: machine learning and natural language processing

Authors: Sarayu Agarwal

Abstract:

This paper aims to teach English (secondary language) by bridging the understanding between the Regional language (primary language) and the English Language (secondary language). Here primary language is the one a person has learned from birth or within the critical period, while secondary language would be any other language one learns or speaks. The paper also focuses on evolving old teaching methods to a contemporary participatory model of learning and teaching. Pilot studies were conducted to gauge an understanding of student’s knowledge of the English language. Teachers and students were interviewed and their academic curriculum was assessed as a part of the initial study. Extensive literature study and design thinking principles were used to devise a solution to the problem. The objective is met using a holistic learning kit/card game to teach children word recognition, word pronunciation, word spelling and writing words. Implication of the paper is a noticeable improvement in the understanding and grasping of English language. With increasing usage and applicability of English as a second language (ESL) world over, the paper becomes relevant due to its easy replicability to any other primary or secondary language. Future scope of this paper would be transforming the idea of participatory learning into self-regulated learning methods. With the upcoming govt. learning centres in rural areas and provision of smart devices such as tablets, the development of the card games into digital applications seems very feasible.

Keywords: English as a second language, vocabulary-building card games, learning through gamification, rural education

Procedia PDF Downloads 235

Authors: Suwaree Yordchim

Abstract:

The objective is to study the satisfaction on English with an online learning. Online learning system mainly consists of English lessons, exercises, tests, web boards, and supplementary lessons for language practice. The sample groups are 80 Thai students studying English for Business Communication, majoring in Hotel and Lodging Management. The data are analyzed by mean, standard deviation (S.D.) value from the questionnaires. The results were found that the most average of satisfaction on academic aspects are technological searching tool through E-learning system that support the students’ learning (4.51), knowledge evaluation on prepost learning and teaching (4.45), and change for project selections according to their interest, subject contents including practice in the real situations (4.45), respectively.

Keywords: English language learning, online system, online learning, supplementary lessons

Procedia PDF Downloads 449

Authors: Talal Alshammari, Nasser Alshammari, Mohamed Sedky, Chris Howard

Abstract:

With the widespread adoption of the Internet-connected devices, and with the prevalence of the Internet of Things (IoT) applications, there is an increased interest in machine learning techniques that can provide useful and interesting services in the smart home domain. The areas that machine learning techniques can help advance are varied and ever-evolving. Classifying smart home inhabitants’ Activities of Daily Living (ADLs), is one prominent example. The ability of machine learning technique to find meaningful spatio-temporal relations of high-dimensional data is an important requirement as well. This paper presents a comparative evaluation of state-of-the-art machine learning techniques to classify ADLs in the smart home domain. Forty-two synthetic datasets and two real-world datasets with multiple inhabitants are used to evaluate and compare the performance of the identified machine learning techniques. Our results show significant performance differences between the evaluated techniques. Such as AdaBoost, Cortical Learning Algorithm (CLA), Decision Trees, Hidden Markov Model (HMM), Multi-layer Perceptron (MLP), Structured Perceptron and Support Vector Machines (SVM). Overall, neural network based techniques have shown superiority over the other tested techniques.

Keywords: activities of daily living, classification, internet of things, machine learning, prediction, smart home

Procedia PDF Downloads 341

Authors: Sheryl Bernardo-Hinesley

Abstract:

Heritage language learners are students who have been raised in a home where a minority language is spoken, who speaks or merely understand the minority heritage language, but to some degree are bilingual in the majority and the heritage language. In view of the rising university enrollment by Hispanics in the United States who have chosen to study Spanish, university language programs are currently faced with challenges of accommodating the language needs of heritage language learners of Spanish. The present study investigates the heritage language perception and language attitudes by heritage language learners of Spanish, as well as their classroom language learning experiences and needs. In order to carry out the study, a qualitative survey was used to gather data from university students. Analysis of students' responses indicates that heritage learners are motivated to learn the heritage language. In relation to the aspects of focus of a language course for heritage learners, results show that the aspects of interest are accent marks and spelling, grammatical accuracy, vocabulary, writing, reading, and culture.

Keywords: heritage language learners, language acquisition, linguistic attitudes, Spanish in the US

Procedia PDF Downloads 196

Authors: Jairson Barbosa Rodrigues, Paulo Romero Martins Maciel, Germano Crispim Vasconcelos

Abstract:

This paper presents an investigation of the performance impacts regarding the variation of five factors (input data size, node number, cores, memory, and disks) when applying a distributed implementation of Naïve Bayes for text classification of a large Corpus on the Spark big data processing framework. Problem: The algorithm's performance depends on multiple factors, and knowing before-hand the effects of each factor becomes especially critical as hardware is priced by time slice in cloud environments. Objectives: To explain the functional relationship between factors and performance and to develop linear predictor models for time and cost. Methods: the solid statistical principles of Design of Experiments (DoE), particularly the randomized two-level fractional factorial design with replications. This research involved 48 real clusters with different hardware arrangements. The metrics were analyzed using linear models for screening, ranking, and measurement of each factor's impact. Results: Our findings include prediction models and show some non-intuitive results about the small influence of cores and the neutrality of memory and disks on total execution time, and the non-significant impact of data input scale on costs, although notably impacts the execution time.

Keywords: big data, design of experiments, distributed machine learning, natural language processing, spark

Procedia PDF Downloads 102

Authors: Elham Sharifineyestani, Mohammad Farshchin

Abstract:

Tropical cyclones are one of the most expensive and deadliest natural disasters. They cause heavy rainfall and serious flash flooding that result in billions of dollars of damage and considerable mortality each year in the United States. Prediction of the frequency of tropical cyclone-induced rainfall events can be helpful in emergency planning and flood risk management. In this study, we have developed a machine-learning model to predict the exceedance frequencies of tropical cyclone-induced rainfall events in the United States. Model results show a satisfactory agreement with available observations. To examine the effectiveness of our approach, we also have compared the result of our predictions with the exceedance frequencies predicted using a physics-based rainfall model by Feldmann.

Keywords: flash flooding, tropical cyclones, frequencies, machine learning, risk management

Procedia PDF Downloads 232

Authors: Ng Liang Shen, Hau Yuan Wen

Abstract:

Arrhythmia analysis of ECG signal plays a major role in diagnosing most of the cardiac diseases. Therefore, a single arrhythmia detection of an electrocardiographic (ECG) record can determine multiple pattern of various algorithms and match accordingly each ECG beats based on Machine Learning supervised learning. These researchers used different features and classification methods to classify different arrhythmia types. A major problem in these studies is the fact that the symptoms of the disease do not show all the time in the ECG record. Hence, a successful diagnosis might require the manual investigation of several hours of ECG records. The point of this paper presents investigations cardiovascular ailment in Electrocardiogram (ECG) Signals for Cardiac Arrhythmia utilizing examination of ECG irregular wave frames via heart beat as correspond arrhythmia which with Machine Learning Pattern Recognition.

Keywords: electrocardiogram, ECG, classification, machine learning, pattern recognition, detection, QRS

Procedia PDF Downloads 357

Authors: Andrei Nosov

Abstract:

This paper discusses the development directions of dialogue systems within the digitalisation of the railway industry, where technologies based on conversational AI are already potentially applied or will be applied. Conversational AI is one of the popular natural language processing (NLP) tasks, as it has great prospects for real-world applications today. At the same time, it is a challenging task as it involves many areas of NLP based on complex computations and deep insights from linguistics and psychology. In this review, we focus on dialogue systems and their implementation in the railway domain. We comprehensively review the state-of-the-art research results on dialogue systems and analyse them from three perspectives: type of problem to be solved, type of model, and type of system. In particular, from the perspective of the type of tasks to be solved, we discuss characteristics and applications. This will help to understand how to prioritise tasks. In terms of the type of models, we give an overview that will allow researchers to become familiar with how to apply them in dialogue systems. By analysing the types of dialogue systems, we propose an unconventional approach in contrast to colleagues who traditionally contrast goal-oriented dialogue systems with open-domain systems. Our view focuses on considering retrieval and generative approaches. Furthermore, the work comprehensively presents evaluation methods and datasets for dialogue systems in the railway domain to pave the way for future research. Finally, some possible directions for future research are identified based on recent research results.

Keywords: digitalisation, railway, dialogue systems, conversational AI, natural language processing, natural language understanding, natural language generation

Procedia PDF Downloads 53

Authors: Kanika Gupta, Ashok Kumar

Abstract:

Biofilms are dense, highly hydrated cell clusters that are irreversibly attached to a substratum, to an interface or to each other, and are embedded in a self-produced gelatinous matrix composed of extracellular polymeric substances. Research in biofilm field has become very significant, as biofilm has shown high mechanical resilience and resistance to antibiotic treatment and constituted as a significant problem in both healthcare and other industry related to microorganisms. The massive information both stated and hidden in the biofilm literature are growing exponentially therefore it is not possible for researchers and practitioners to automatically extract and relate information from different written resources. So, the current work proposes and discusses the use of text mining techniques for the extraction of information from biofilm literature corpora containing 34306 documents. It is very difficult and expensive to obtain annotated material for biomedical literature as the literature is unstructured i.e. free-text. Therefore, we considered unsupervised approach, where no annotated training is necessary and using this approach we developed a system that will classify the text on the basis of growth and development, drug effects, radiation effects, classification and physiology of biofilms. For this, a two-step structure was used where the first step is to extract keywords from the biofilm literature using a metathesaurus and standard natural language processing tools like Rapid Miner_v5.3 and the second step is to discover relations between the genes extracted from the whole set of biofilm literature using pubmed.mineR_v1.0.11. We used unsupervised approach, which is the machine learning task of inferring a function to describe hidden structure from 'unlabeled' data, in the above-extracted datasets to develop classifiers using WinPython-64 bit_v3.5.4.0Qt5 and R studio_v0.99.467 packages which will automatically classify the text by using the mentioned sets. The developed classifiers were tested on a large data set of biofilm literature which showed that the unsupervised approach proposed is promising as well as suited for a semi-automatic labeling of the extracted relations. The entire information was stored in the relational database which was hosted locally on the server. The generated biofilm vocabulary and genes relations will be significant for researchers dealing with biofilm research, making their search easy and efficient as the keywords and genes could be directly mapped with the documents used for database development.

Keywords: biofilms literature, classifiers development, text mining, unsupervised learning approach, unstructured data, relational database

Procedia PDF Downloads 155

Authors: Vishnu Pratap Singh Kirar, Madhuri Saxena

Abstract:

Advancement in sequence data generation technologies is churning out voluminous omics data and posing a massive challenge to annotate the biological functional features. With so much data available, the use of machine learning methods and tools to make novel inferences has become obvious. Machine learning methods have been successfully applied to a lot of disciplines, including computational biology and bioinformatics. Researchers in computational biology are interested to develop novel machine learning frameworks to classify the huge amounts of biological data. In this proposal, it plan to employ novel machine learning approaches to aid the understanding of how apparently innocuous mutations (in intergenic DNA and at synonymous sites) cause diseases. We are also interested in discovering novel functional sites in the genome and mutations in which can affect a phenotype of interest.

Keywords: genome wide association studies (GWAS), next generation sequencing (NGS), deep learning, omics

Procedia PDF Downloads 81

Authors: Shuen-Tai Wang, Fang-An Kuo, Chau-Yi Chou, Yu-Bin Fang

Abstract:

2016 has become the year of the Artificial Intelligence explosion. AI technologies are getting more and more matured that most world well-known tech giants are making large investment to increase the capabilities in AI. Machine learning is the science of getting computers to act without being explicitly programmed, and deep learning is a subset of machine learning that uses deep neural network to train a machine to learn  features directly from data. Deep learning realizes many machine learning applications which expand the field of AI. At the present time, deep learning frameworks have been widely deployed on servers for deep learning applications in both academia and industry. In training deep neural networks, there are many standard processes or algorithms, but the performance of different frameworks might be different. In this paper we evaluate the running performance of two state-of-the-art distributed deep learning frameworks that are running training calculation in parallel over multi GPU and multi nodes in our cloud environment. We evaluate the training performance of the frameworks with ResNet-50 convolutional neural network, and we analyze what factors that result in the performance among both distributed frameworks as well. Through the experimental analysis, we identify the overheads which could be further optimized. The main contribution is that the evaluation results provide further optimization directions in both performance tuning and algorithmic design.

Keywords: artificial intelligence, machine learning, deep learning, convolutional neural networks

Procedia PDF Downloads 192

Authors: Amy Hodges, Deanna Rasmussen, Sherry Ward

Abstract:

This paper examines the use of ePortfolios in a two-course sequence for ESL (English as a Second Language) students at an international branch campus in Doha, Qatar. ePortfolios support the transfer of language learning, but few have examined the sustainability of that transfer across an ESL program. Drawing upon surveys and interviews with students, we analyze three case studies that complicate previous research on metacognition, language learning, and ePortfolios. Our findings have implications for those involved in ESL programs and assessment of student writing.

Keywords: TESOL, electronic portfolios, assessment, technology

Procedia PDF Downloads 249

Authors: L. Freeborn

Abstract:

Recent years have seen an increasing number of neuroimaging studies related to language learning as imaging techniques such as fMRI and EEG have become more widely accessible to researchers. By using a variety of structural and functional neuroimaging techniques, these studies have already made considerable progress in terms of our understanding of neural networks and processing related to first and second language acquisition. However, the methodological designs employed in neuroimaging studies to test language learning have been questioned by applied linguists working within the field of second language acquisition (SLA). One of the major criticisms is that tasks designed to measure language learning gains rarely have a communicative function, and seldom assess learners’ ability to use the language in authentic situations. This brings the validity of many neuroimaging tasks into question. The fundamental reason why people learn a language is to communicate, and it is well-known that both first and second language proficiency are developed through meaningful social interaction. With this in mind, the SLA field is in agreement that second language acquisition and proficiency should be measured through learners’ ability to communicate in authentic real-life situations. Whilst authenticity is not always possible to achieve in a classroom environment, the importance of task authenticity should be reflected in the design of language assessments, teaching materials, and curricula. Tasks that bear little relation to how language is used in real-life situations can be considered to lack construct validity. This paper first describes the typical tasks used in neuroimaging studies to measure language gains and proficiency, then analyses to what extent these tasks can validly assess these constructs.

Keywords: neuroimaging studies, research design, second language acquisition, task validity

Procedia PDF Downloads 119

Authors: Hailyie Tekleselase

Abstract:

Deep learning is increasingly used as a building block of security systems. However, neural networks are hard to interpret and typically solid to the practitioner. This paper presents a detail survey of computing methods in cyber security, and analyzes the prospects of enhancing the cyber security capabilities by suggests that of accelerating the intelligence of the security systems. There are many AI-based applications used in industrial scenarios such as Internet of Things (IoT), smart grids, and edge computing. Machine learning technologies require a training process which introduces the protection problems in the training data and algorithms. We present machine learning techniques currently applied to the detection of intrusion, malware, and spam. Our conclusions are based on an extensive review of the literature as well as on experiments performed on real enterprise systems and network traffic. We conclude that problems can be solved successfully only when methods of artificial intelligence are being used besides human experts or operators.

Keywords: artificial intelligence, machine learning, deep learning, cyber security, big data

Procedia PDF Downloads 113

Authors: Du T. Tran, Quyen T. L. Hoang

Abstract:

The role of language learning strategies in second language acquisition has received increased attention across several disciplines in recent years. Language learning strategies have been shown to occur in many studies over the passing years with the aim of improving the efficiency of language learning. Following previous studies, this study endeavors to scrutinize language learning strategies employed by the students at Vo Minh Duc high school and the effect of motivation on students’ learning strategy choices. The responses are examined quantitatively and qualitatively to enhance their validity and reliability. Data are collected from 342 students’ responses to the questionnaire, interviews with ten teachers and fifteen students, and classroom observations. The findings reveal that students’ motivation has an enormous impact on the choice of language learning strategies. The results simultaneously show that students use many language learning strategies to enhance their communicative competence, but the most frequently used ones are cognitive and affective ones. Significant correlations among types of learning strategies and the influence of motivation on the choices of language learning strategies were consistent with previous studies. The study’s results are expected to be beneficial to teachers of English and students in terms of narrowing the gap between the students' language learning strategies and their teaching methodologies preferences and sketching out the best strategies to enhance students’ speaking skills. The implications of these findings and the importance of viewing learners holistically are discussed, and recommendations are made for ongoing research.

Keywords: learning strategies, speaking skills, memorization strategies, cognitive strategies, affective strategies

Procedia PDF Downloads 91

Authors: Sakib Hasn, Shahid Anwar

Abstract:

This study highlights the vital importance of assessing the Simulator Sickness Questionnaire and presence measures as virtual reality (VR) incorporation into language teaching gains popularity. To address user discomfort, which prevents efficient learning in VR environments, the measurement of SSQ becomes crucial. Additionally, evaluating presence metrics is essential to determine the level of engagement and immersion, both crucial for rich language learning experiences. This paper designs a VR-based Chinese language application and proposes a thorough test technique aimed at systematically analyzing SSQ and presence measures. Subjective tests and data analysis were carried out to highlight the significance of addressing user discomfort in VR language education. The results of this study shed light on the difficulties posed by user discomfort in VR language learning and offer insightful advice on how to improve VR language learning applications. Furthermore, the outcome of the research explores ‘VR-based language education,’ ‘inclusive language learning platforms," and "cross-cultural communication,’ highlighting the potential for VR to facilitate language learning across diverse cultural backgrounds. Overall, the analysis results contribute to the enrichment of language learning experiences in the virtual realm and underscore the need for continued exploration and improvement in this field.

Keywords: virtual reality (VR), language education, simulator sickness questionnaire, presence metrics, VR-based Chinese language education

Procedia PDF Downloads 48

Authors: Jun Wang, Ge Zhang

Abstract:

Machine learning has been used in many areas today. It thrives at reviewing large volumes of data and identifying patterns and trends that might not be apparent to a human. Given the huge potential benefit and the amount of data available in the financial market, it is not surprising to see machine learning applied to various financial products. While future prices of financial securities are extremely difficult to forecast, we study them from a different angle. Instead of trying to forecast future prices, we apply machine learning algorithms to predict the direction of future price movement, in particular, whether a sector Exchange Traded Fund (ETF) would outperform or underperform the market in the next week or in the next month. We apply several machine learning algorithms for this prediction. The algorithms are Linear Discriminant Analysis (LDA), k-Nearest Neighbors (KNN), Decision Tree (DT), Gaussian Naive Bayes (GNB), and Neural Networks (NN). We show that these machine learning algorithms, most notably GNB and NN, have some predictive power in forecasting out-performance and under-performance out of sample. We also try to explore whether it is possible to utilize the predictions from these algorithms to outperform the buy-and-hold strategy of the S&P 500 index. The trading strategy to explore out-performance predictions does not perform very well, but the trading strategy to explore under-performance predictions can earn higher returns than simply holding the S&P 500 index out of sample.

Keywords: machine learning, ETF prediction, dynamic trading, asset allocation

Procedia PDF Downloads 77

Authors: Abrar Al-Hasan

Abstract:

Web 2.0 has significantly increased the amount of information available to users not only about firms and their offerings, but also about the activities of other individuals in their networks and markets. It is widely acknowledged that this increased availability of ‘social’ information, particularly about other individuals, is likely to influence a user’s behavior and choices. However, there are very few systematic studies of how such increased information transparency on the behavior of other users in a focal users’ network influences a focal users’ behavior in the emerging marketplace of online language learning. This study seeks to examine the value and impact of ‘social activities’ – wherein, a user sees and interacts with the learning activities of her peers – on her language learning efficiency. An online experiment in a peer-to-peer language marketplace was conducted to compare the learning efficiency of users with ‘social’ information versus users with no ‘social’ information. The results of this study highlight the impact and importance of ‘social’ information within the language learning context. The study concludes by exploring how these insights may inspire new developments in online education.

Keywords: e-Learning, language learning marketplace, peer-to-peer, social network

Procedia PDF Downloads 371

Authors: John McNamara

Abstract:

Growing up creative is an important idea in today’s classrooms. As education seeks to prepare children for their futures, it is important that the system considers traditional as well as non-traditional pathways. This poster describes the findings of a research study investigating creative potential in children with learning disabilities. Children with learning disabilities were administered the Torrance Test of Creative Problem Solving along with subtests from the Comprehensive Test of Phonological Processing. A quantitative comparative analysis was computed using paired-sample t-tests. Results indicated statistically significant difference between children’s creative problem-solving skills and their reading-based skills. The results lend support to the idea that children with learning disabilities have inherent strengths in the area of creativity. It can be hypothesized that the success of these children may be associated with the notion that they are using a type of neurological processing that is not otherwise used in academic tasks. Children with learning disabilities, a presumed left-side neurological processing problem, process information with the right side of the brain – even with tasks that should be processed with the left side (i.e. language). In over-using their right hemisphere, it is hypothesized that children with learning disabilities have well-developed right hemispheres and, as such, have strengths associated with this type of processing, such as innovation and creativity. The current study lends support to the notion that children with learning disabilities may be particularly primed to succeed in areas that call on creativity and creative thinking.

Keywords: learning disabilities, educational psychology, education, creativity

Procedia PDF Downloads 63

Authors: Ozlem Bozok, Yusuf Bozok

Abstract:

The importance of using mother tongue and translation in foreign language classrooms cannot be ignored and translation can be utilized as a method in English Language Teaching courses. There exist researches advocating or objecting to the use of translation in foreign language learning but they all have a point in common: Translation should be used as an aid to teaching, not an end in itself. In this research, prospective English language teachers’ opinions about translation use and use of mother tongue in foreign language teaching are investigated and according to the findings, some explanations and recommendations are made.

Keywords: exposure to foreign language translation, foreign language learning, prospective teachers’ opinions, use of L1

Procedia PDF Downloads 518

Authors: Rachid Ed-Dali, Khalid Elasri

Abstract:

Mass media seem to provide a rich content for language acquisition. Exposure to television, the Internet, the mobile phone and other technological gadgets and devices helps enrich the student’s lexicon positively as well as negatively. The difficulties encountered by students while learning and acquiring second languages in addition to their eagerness to comprehend the content of a particular program prompt them to diversify their methods so as to achieve their targets. The present study highlights the significance of certain media channels and their involvement in language acquisition with the employment of the Natural Approach to further grasp whether students, especially secondary and high school students, learn and acquire errors through watching subtitled television programs. The chief objective is investigating the deductive and inductive relevance of certain programs beside the involvement of peripheral learning while acquiring mistakes.

Keywords: errors, mistakes, Natural Approach, peripheral learning, solecism

Procedia PDF Downloads 108

Authors: Senol Dogan, Gunay Karli

Abstract:

Identification of relevant publications based on a Medline query is time-consuming and error-prone. An all based process has the potential to solve this problem without any manual work. To the best of our knowledge, our study is the first to investigate the ability of machine learning to identify relevant articles accurately. 5 different machine learning algorithms were tested using 23 predictors based on several metadata fields attached to publications. We find that the Boosted model is the best-performing algorithm and its overall accuracy is 96%. In addition, specificity and sensitivity of the algorithm is 97 and 93%, respectively. As a result of the work, we understood that we can apply the same procedure to understand cancer gene expression big data.

Keywords: prediction of publications, machine learning, algorithms, bariatric surgery, comparison of algorithms, boosted, tree, logistic regression, ANN model

Procedia PDF Downloads 196

Authors: Christian Neunzig, Simon Fahle, Jürgen Schulz, Matthias Möller, Bernd Kuhlenkötter

Abstract:

Due to increasing cost pressure in global markets, artificial intelligence is becoming a technology that is decisive for competition. Predictive quality enables machinery and plant manufacturers to ensure product quality by using data-driven forecasts via machine learning models as a decision-making basis for test results. The use of cross-process Bosch production data along the value chain of hydraulic valves is a promising approach to classifying the quality characteristics of workpieces.

Keywords: predictive quality, hydraulics, machine learning, classification, supervised learning

Procedia PDF Downloads 218

Authors: Pi-Lan Yang

Abstract:

It is controversial which type of gloss condition (i.e., gloss language or gloss position) is more effective in second or foreign language (L2) vocabulary learning. The present study compared the performance on learning ten English words in the conditions of L2 English reading with no glosses and with glosses of Chinese equivalents/translations and L2 English definitions at the side of a page and at an attached sheet for low-intermediate Chinese-speaking learners of English, who were technology university students in Taiwan. It is found first that the performances on the immediate posttest and the delayed posttest were overall better in the gloss condition than those in the no-gloss condition. Next, it is found that the glosses of Chinese translations were more effective and sustainable than those of L2 English definitions. Finally, the effects of L2 English glosses at the side of a page were observed to be less sustainable than those at an attached sheet. In addition, an opinion questionnaire used also showed a preference for the glosses of Chinese translations in L2 English reading. These results would be discussed in terms of automated lexical access, sentence processing mechanisms, and the trade-off nature of storage and processing functions in working memory system, proposed by the capacity theory of language comprehension.

Keywords: glosses of Chinese equivalents/translations, glosses of L2 English definitions, L2 vocabulary learning, L2 English reading

Procedia PDF Downloads 235

Authors: Bowen Ding, Yihao Kuang

Abstract:

In recent years, large-scale pre-trained language models have achieved state-of-the-art performance on a variety of natural language processing tasks. The word vectors produced by these language models can be viewed as dense encoded presentations of natural language that in text form. However, it is unknown how much linguistic information is encoded and how. In this paper, we construct several corresponding probing tasks for multiple linguistic information to clarify the encoding capabilities of different language models and performed a visual display. We firstly obtain word presentations in vector form from different language models, including BERT, ELMo, RoBERTa and GPT. Classifiers with a small scale of parameters and unsupervised tasks are then applied on these word vectors to discriminate their capability to encode corresponding linguistic information. The constructed probe tasks contain both semantic and syntactic aspects. The semantic aspect includes the ability of the model to understand semantic entities such as numbers, time, and characters, and the grammatical aspect includes the ability of the language model to understand grammatical structures such as dependency relationships and reference relationships. We also compare encoding capabilities of different layers in the same language model to infer how linguistic information is encoded in the model.

Keywords: language models, probing task, text presentation, linguistic information

Procedia PDF Downloads 92

Authors: Junjie Zhong, Kai Hong, Lei Wang

Abstract:

Passage Retrieval (PR) plays an important role in many Natural Language Processing (NLP) tasks. Traditional efficient retrieval models relying on exact term-matching, such as TF-IDF or BM25, have nowadays been exceeded by pre-trained language models which match by semantics. Though they gain effectiveness, deep language models often require large memory as well as time cost. To tackle the trade-off between efficiency and effectiveness in PR, this paper proposes Graph Passage Retriever (GraphPR), a graph-based model inspired by the development of graph learning techniques. Different from existing works, GraphPR is end-to-end and integrates both term-matching information and semantics. GraphPR constructs a passage-level graph from BM25 retrieval results and trains a GCN-like model on the graph with graph-based objectives. Passages were regarded as nodes in the constructed graph and were embedded in dense vectors. PR can then be implemented using embeddings and a fast vector-similarity search. Experiments on a variety of real-world retrieval datasets show that the proposed model outperforms related models in several evaluation metrics (e.g., mean reciprocal rank, accuracy, F1-scores) while maintaining a relatively low query latency and memory usage.

Keywords: efficiency, effectiveness, graph learning, language model, passage retrieval, term-matching model

Procedia PDF Downloads 120

Authors: Stefan Papastefanou

Abstract:

Artificial intelligence (AI) is an interdisciplinary field of computer science with the aim of creating intelligent machine behavior. Early approaches to AI have been configured to operate in very constrained environments where the behavior of the AI system was previously determined by formal rules. Knowledge was presented as a set of rules that allowed the AI system to determine the results for specific problems; as a structure of if-else rules that could be traversed to find a solution to a particular problem or question. However, such rule-based systems typically have not been able to generalize beyond the knowledge provided. All over the world and especially in IT-heavy industries such as the United States, the European Union, Singapore, and China, machine learning has developed to be an immense asset, and its applications are becoming more and more significant. It has to be examined how such products of machine learning models can and should be protected by IP law and for the purpose of this paper patent law specifically, since it is the IP law regime closest to technical inventions and computing methods in technical applications. Genetic breeding models are currently less popular than recursive neural network method and deep learning, but this approach can be more easily described by referring to the evolution of natural organisms, and with increasing computational power; the genetic breeding method as a subset of the evolutionary algorithms models is expected to be regaining popularity. The research method focuses on patentability (according to the world’s most significant patent law regimes such as China, Singapore, the European Union, and the United States) of AI inventions and machine learning. Questions of the technical nature of the problem to be solved, the inventive step as such, and the question of the state of the art and the associated obviousness of the solution arise in the current patenting processes. Most importantly, and the key focus of this paper is the problem of patenting inventions that themselves are developed through machine learning. The inventor of a patent application must be a natural person or a group of persons according to the current legal situation in most patent law regimes. In order to be considered an 'inventor', a person must actually have developed part of the inventive concept. The mere application of machine learning or an AI algorithm to a particular problem should not be construed as the algorithm that contributes to a part of the inventive concept. However, when machine learning or the AI algorithm has contributed to a part of the inventive concept, there is currently a lack of clarity regarding the ownership of artificially created inventions. Since not only all European patent law regimes but also the Chinese and Singaporean patent law approaches include identical terms, this paper ultimately offers a comparative analysis of the most relevant patent law regimes.

Keywords: algorithms, inventor, genetic breeding models, machine learning, patentability

Procedia PDF Downloads 100

Authors: Tanmay Bisen, Aastha Shayla, Susham Biswas

Abstract:

Oil spills from tankers can cause significant harm to the environment and local communities, as well as have economic consequences. Early predictions of oil spills can help to minimize these impacts. Our proposed system uses machine learning and neural networks to predict potential oil spills by monitoring data from ship Automatic Identification Systems (AIS). The model analyzes ship movements, speeds, and changes in direction to identify patterns that deviate from the norm and could indicate a potential spill. Our approach not only identifies anomalies but also predicts spills before they occur, providing early detection and mitigation measures. This can prevent or minimize damage to the reputation of the company responsible and the country where the spill takes place. The model's performance on the MV Wakashio oil spill provides insight into its ability to detect and respond to real-world oil spills, highlighting areas for improvement and further research.

Keywords: Anomaly Detection, Oil Spill Prediction, Machine Learning, Image Processing, Graph Neural Network (GNN)

Procedia PDF Downloads 55

Authors: Simon Espigolé, Igor Shalyminov, Helen Hastie

Abstract:

Deep machine learning techniques including deep neural networks (DNN) have been used to model language and dialogue for conversational agents to perform tasks, such as giving technical support and also for general chit-chat. They have been shown to be capable of generating long, diverse and coherent sentences in end-to-end dialogue systems and natural language generation. However, these systems tend to imitate the training data and will only generate the concepts and language within the scope of what they have been trained on. This work explores how deep neural networks can be used in a task that would normally require human creativity, whereby the human would read the movie description and/or watch the movie and come up with a compelling, interesting movie title. This task differs from simple summarization in that the movie title may not necessarily be derivable from the content or semantics of the movie description. Here, we train a type of DNN called a sequence-to-sequence model (seq2seq) that takes as input a short textual movie description and some information on e.g. genre of the movie. It then learns to output a movie title. The idea is that the DNN will learn certain techniques and approaches that the human movie titler may deploy that may not be immediately obvious to the human-eye. To give an example of a generated movie title, for the movie synopsis: ‘A hitman concludes his legacy with one more job, only to discover he may be the one getting hit.’; the original, true title is ‘The Driver’ and the one generated by the model is ‘The Masquerade’. A human evaluation was conducted where the DNN output was compared to the true human-generated title, as well as a number of baselines, on three 5-point Likert scales: ‘creativity’, ‘naturalness’ and ‘suitability’. Subjects were also asked which of the two systems they preferred. The scores of the DNN model were comparable to the scores of the human-generated movie title, with means m=3.11, m=3.12, respectively. There is room for improvement in these models as they were rated significantly less ‘natural’ and ‘suitable’ when compared to the human title. In addition, the human-generated title was preferred overall 58% of the time when pitted against the DNN model. These results, however, are encouraging given the comparison with a highly-considered, well-crafted human-generated movie title. Movie titles go through a rigorous process of assessment by experts and focus groups, who have watched the movie. This process is in place due to the large amount of money at stake and the importance of creating an effective title that captures the audiences’ attention. Our work shows progress towards automating this process, which in turn may lead to a better understanding of creativity itself.

Keywords: creativity, deep machine learning, natural language generation, movies

Procedia PDF Downloads 315

Authors: Adriana Haulica

Abstract:

Powered by Machine Learning, Precise medicine is tailored by now to use genetic and molecular profiling, with the aim of optimizing the therapeutic benefits for cohorts of patients. As the majority of Machine Language algorithms come from heuristics, the outputs have contextual validity. This is not very restrictive in the sense that medicine itself is not an exact science. Meanwhile, the progress made in Molecular Biology, Bioinformatics, Computational Biology, and Precise Medicine, correlated with the huge amount of human biology data and the increase in computational power, opens new healthcare challenges. A more accurate diagnosis is needed along with real-time treatments by processing as much as possible from the available information. The purpose of this paper is to present a deeper vision for the future of Artificial Intelligence in Precise medicine. In fact, actual Machine Learning algorithms use standard mathematical knowledge, mostly Euclidian metrics and standard computation rules. The loss of information arising from the classical methods prevents obtaining 100% evidence on the diagnosis process. To overcome these problems, we introduce MEDICOMPILLS, a new architectural concept tool of information processing in Precise medicine that delivers diagnosis and therapy advice. This tool processes poly-field digital resources: global knowledge related to biomedicine in a direct or indirect manner but also technical databases, Natural Language Processing algorithms, and strong class optimization functions. As the name suggests, the heart of this tool is a compiler. The approach is completely new, tailored for omics and clinical data. Firstly, the intrinsic biological intuition is different from the well-known “a needle in a haystack” approach usually used when Machine Learning algorithms have to process differential genomic or molecular data to find biomarkers. Also, even if the input is seized from various types of data, the working engine inside the MEDICOMPILLS does not search for patterns as an integrative tool. This approach deciphers the biological meaning of input data up to the metabolic and physiologic mechanisms, based on a compiler with grammars issued from bio-algebra-inspired mathematics. It translates input data into bio-semantic units with the help of contextual information iteratively until Bio-Logical operations can be performed on the base of the “common denominator “rule. The rigorousness of MEDICOMPILLS comes from the structure of the contextual information on functions, built to be analogous to mathematical “proofs”. The major impact of this architecture is expressed by the high accuracy of the diagnosis. Detected as a multiple conditions diagnostic, constituted by some main diseases along with unhealthy biological states, this format is highly suitable for therapy proposal and disease prevention. The use of MEDICOMPILLS architecture is highly beneficial for the healthcare industry. The expectation is to generate a strategic trend in Precise medicine, making medicine more like an exact science and reducing the considerable risk of errors in diagnostics and therapies. The tool can be used by pharmaceutical laboratories for the discovery of new cures. It will also contribute to better design of clinical trials and speed them up.

Keywords: bio-semantic units, multiple conditions diagnosis, NLP, omics

Procedia PDF Downloads 58