LKMT:Linguistics Knowledge-Driven Multi-Task Neural Machine Translation for Urdu and English

Thanks to the strong representation capability of pre-trained language models,supervised machine translation models have achieved outstanding performance.However,the performances of these models drop sharply when the scale of the parallel training corpus is limited.Considering the pre-trained language model has a strong ability for monolingual representation,it is the key challenge for machine translation to construct the in-depth relationship between the source and target language by injecting the lexical and syntactic information into pre-trained language models.To alleviate the dependence on the parallel corpus,we propose a Linguistics Knowledge-Driven MultiTask(LKMT)approach to inject part-of-speech and syntactic knowledge into pre-trained models,thus enhancing the machine translation performance.On the one hand,we integrate part-of-speech and dependency labels into the embedding layer and exploit large-scale monolingual corpus to update all parameters of pre-trained language models,thus ensuring the updated language model contains potential lexical and syntactic information.On the other hand,we leverage an extra self-attention layer to explicitly inject linguistic knowledge into the pre-trained language model-enhanced machine translation model.Experiments on the benchmark dataset show that our proposed LKMT approach improves the Urdu-English translation accuracy by 1.97 points and the English-Urdu translation accuracy by 2.42 points,highlighting the effectiveness of our LKMT framework.Detailed ablation experiments confirm the positive impact of part-of-speech and dependency parsing on machine translation.

Discrimination of Motor Imagery Patterns by Electroencephalogram Phase Synchronization Combined With Frequency Band Energy

Central nerve signal evoked by thoughts can be directly used to control a robot or prosthetic devices without the involvement of the peripheral nerve and muscles.This is a new strategy of human-computer interaction.A method of electroencephalogram(EEG) phase synchronization combined with band energy was proposed to construct a feature vector for pattern recognition of brain-computer interaction based on EEG induced by motor imagery in this paper,rhythm and beta rhythm were first extracted from EEG by band pass filter and then the frequency band energy was calculated by the sliding time window;the instantaneous phase values were obtained using Hilbert transform and then the phase synchronization feature was calculated by the phase locking value(PLV) and the best time interval for extracting the phase synchronization feature was searched by the distribution of the PLV value in the time domain.Finally,discrimination of motor imagery patterns was performed by the support vector machine(SVM).The results showed that the phase synchronization feature more effective in4s-7s and the correct classification rate was 91.4%.Compared with the results achieved by a single EEG feature related to motor imagery,the correct classification rate was improved by 3.5 and4.3 percentage points by combining phase synchronization with band energy.These indicate that the proposed method is effective and it is expected that the study provides a way to improve the performance of the online real-time brain-computer interaction control system based on EEG related to motor imagery.

Review Expert Collaborative Recommendation Algorithm Based on Topic Relationship

The project review information plays an important role in the recommendation of review experts. In this paper, we aim to determine review expert's rating by using the historical rating records and the final decision results on the previous projects, and by means of some rules, we construct a rating matrix for projects and experts. For the data sparseness problem of the rating matrix and the 'cold start' problem of new expert recommendation, we assume that those projects/experts with similar topics have similar feature vectors and propose a review expert collaborative recommendation algorithm based on topic relationship. Firstly, we obtain topics of projects/experts based on latent Dirichlet allocation (LDA) model, and build the topic relationship network of projects/experts. Then, through the topic relationship between projects/experts, we find a neighbor collection which shares the largest similarity with target project/expert, and integrate the collection into the collaborative filtering recommendation algorithm based on matrix factorization. Finally, by learning the rating matrix to get feature vectors of the projects and experts, we can predict the ratings that a target project will give candidate review experts, and thus achieve the review expert recommendation. Experiments on real data set show that the proposed method could predict the review expert rating more effectively, and improve the recommendation effect of review experts. © 2014 Chinese Association of Automation.

Text Detection in Natural Scene Images Using Morphological Component Analysis and Laplacian Dictionary

Text in natural scene images usually carries abundant semantic information. However, due to variations of text and complexity of background, detecting text in scene images becomes a critical and challenging task. In this paper, we present a novel method to detect text from scene images. Firstly, we decompose scene images into background and text components using morphological component analysis(MCA), which will reduce the adverse effects of complex backgrounds on the detection results.In order to improve the performance of image decomposition,two discriminative dictionaries of background and text are learned from the training samples. Moreover, Laplacian sparse regularization is introduced into our proposed dictionary learning method which improves discrimination of dictionary. Based on the text dictionary and the sparse-representation coefficients of text, we can construct the text component. After that, the text in the query image can be detected by applying certain heuristic rules. The results of experiments show the effectiveness of the proposed method.

Protein-Protein Interaction Extraction Based on Convex Combination Kernel Function

Owing to the effect of classified models was different in Protein-Protein Interaction(PPI) extraction, which was made by different single kernel functions, and only using single kernel function hardly trained the optimal classified model to extract PPI, this paper presents a strategy to find the optimal kernel function from a kernel function set. The strategy is that in the kernel function set which consists of different single kernel functions, endlessly finding the last two kernel functions on the performance in PPI extraction, using their optimal kernel function to replace them, until there is only one kernel function and it’s the final optimal kernel function. Finally, extracting PPI using the classified model made by this kernel function. This paper conducted the PPI extraction experiment on AIMed corpus, the experimental result shows that the optimal convex combination kernel function this paper presents can effectively improve the extraction performance than single kernel function, and it gets the best precision which reaches 65.0 among the similar PPI extraction systems.

Enhancing low-resource cross-lingual summarization from noisy data with fine-grained reinforcement learning

Cross-lingual summarization(CLS)is the task of generating a summary in a target language from a document in a source language.Recently,end-to-end CLS models have achieved impressive results using large-scale,high-quality datasets typically constructed by translating monolingual summary corpora into CLS corpora.However,due to the limited performance of low-resource language translation models,translation noise can seriously degrade the performance of these models.In this paper,we propose a fine-grained reinforcement learning approach to address low-resource CLS based on noisy data.We introduce the source language summary as a gold signal to alleviate the impact of the translated noisy target summary.Specifically,we design a reinforcement reward by calculating the word correlation and word missing degree between the source language summary and the generated target language summary,and combine it with cross-entropy loss to optimize the CLS model.To validate the performance of our proposed model,we construct Chinese-Vietnamese and Vietnamese-Chinese CLS datasets.Experimental results show that our proposed model outperforms the baselines in terms of both the ROUGE score and BERTScore.

Semantic-aware entity alignment for low resource language knowledge graph

Entity alignment(EA)is an important technique aiming to find the same real entity between two different source knowledge graphs(KGs).Current methods typically learn the embedding of entities for EA from the structure of KGs for EA.Most EA models are designed for rich-resource languages,requiring sufficient resources such as a parallel corpus and pre-trained language models.However,low-resource language KGs have received less attention,and current models demonstrate poor performance on those low-resource KGs.Recently,researchers have fused relation information and attributes for entity representations to enhance the entity alignment performance,but the relation semantics are often ignored.To address these issues,we propose a novel Semantic-aware Graph Neural Network(SGNN)for entity alignment.First,we generate pseudo sentences according to the relation triples and produce representations using pre-trained models.Second,our approach explores semantic information from the connected relations by a graph neural network.Our model captures expanded feature information from KGs.Experimental results using three low-resource languages demonstrate that our proposed SGNN approach out performs better than state-of-the-art alignment methods on three proposed datasets and three public datasets.

Single-trial decoding of imagined grip force parameters involving the right or left hand based on movement-related cortical potentials

Time–domain feature representation for imagined grip force movement-related cortical potentials(MRCP)of the right or left hand and the decoding of imagined grip force parameters based on electroencephalogram(EEG)activity recorded during a single trial were here investigated.EEG signals were acquired from eleven healthy subjects during four different imagined tasks performed with the right or left hand.Subjects were instructed to execute imagined grip movement at two different levels of force.Each task was executed 60 times in random order.The imagined grip force MRCP of the right or left hand was analyzed by superposition and averaging technology,a single-trial extraction method,analysis of variance(ANOVA),and multiple comparisons.Significantly different features were observed among different imagined grip force tasks.These differences were used to decode imagined grip force parameters using Fisher linear discrimination analysis based on kernel function(k-FLDA)and support vector machine(SVM).Under the proposed experimental paradigm,the study showed that MRCP may characterize the dynamic processing that takes place in the brain during the planning,execution,and precision of a given imagined grip force task.This means that features related to MRCP can be used to decode imagined grip force parameters based on EEG.ANOVA and multiple comparisons of time–domain features for MRCP showed that movement-monitoring potentials(MMP)and specific interval(0–150 ms)average potentials to be significantly different among 4 different imagined grip force tasks.The minimum peak negativity differed significantly between high and low amplitude grip force.Identification of the 4different imagined grip force tasks based on MMP was performed using k-FLDA and SVM,and the average misclassification rates of 27%±5%and 24%±4%across 11 subjects were achieved respectively.The minimum misclassification rate was 15%,and the average minimum misclassification rate across 11 subjects was24%±4.5%.This investigation indicates that imagined grip force MRCP may encode imagined grip force parameters.Single-trial decoding of imagined grip force parameters based on MRCP may be feasible.The study may provide some additional and fine control instructions for brain–computer interfaces.

Exploiting comments information to improve legal public opinion news abstractive summarization

Automatically generating a brief summary for legal-related public opinion news(LPO-news,which contains legal words or phrases)plays an important role in rapid and effective public opinion disposal.For LPO-news,the critical case elements which are significant parts of the summary may be mentioned several times in the reader comments.Consequently,we investigate the task of comment-aware abstractive text summarization for LPO-news,which can generate salient summary by learning pivotal case elements from the reader comments.In this paper,we present a hierarchical comment-aware encoder(HCAE),which contains four components:1)a traditional sequenceto-sequence framework as our baseline;2)a selective denoising module to filter the noisy of comments and distinguish the case elements;3)a merge module by coupling the source article and comments to yield comment-aware context representation;4)a recoding module to capture the interaction among the source article words conditioned on the comments.Extensive experiments are conducted on a large dataset of legal public opinion news collected from micro-blog,and results show that the proposed model outperforms several existing state-of-the-art baseline models under the ROUGE metrics.

Linguistic feature template integration for Chinese-Vietnamese neural machine translation

1 Introduction and main contributions Template-based approaches have achieved significant progress in low-resource neural machine translation(NMT)recently[1],such as the efficient works,NMT-GTM[2],SoftPrototype[3],etc.However,most previous works only retrieve target sentence as template to generate translation,neglecting the utilization of linguistic feature that contained in the source sentence and template.