Towards Self-supervised Learning for Multi-function Radar Behavior State Detection and Recognition

<div>The analysis of intercepted multi-function radar (MFR) signals has gained considerable attention in the field of cognitive electronic reconnaissance. With the rapid development of MFR, the switch between different work modes is becoming more flexible, increasing the agility of pulse parameters. Most of the existing approaches for recognizing MFR behaviors heavily depend on prior information, which can hardly be obtained in a non-cooperative way. This study develops a novel hierarchical contrastive self-supervise-based method for segmenting and clustering MFR pulse sequences. First, a convolutional neural network (CNN) with a limited receptive field is trained in a contrastive way to distinguish between pulse descriptor words (PDW) in the original order and the samples created by random permutations to detect the boundary between each radar word and perform segmentation. Afterward, the K-means++ algorithm with cosine distances is established to cluster the segmented PDWs according to the output vectors of the CNN’s last layer for radar words extraction. This segmenting and clustering process continues to go in the extracted radar word sequence, radar phase sequence, and so on, finishing the automatic extraction of MFR behavior states in the MFR hierarchical model. Simulation results show that without using any labeled data, the proposed method can effectively mine distinguishable patterns in the sequentially arriving PDWs and recognize the MFR behavior states under corrupted, overlapped pulse parameters.</div>

Towards Self-supervised Learning for Multi-function Radar Behavior State Detection and Recognition

<div>The analysis of intercepted multi-function radar (MFR) signals has gained considerable attention in the field of cognitive electronic reconnaissance. With the rapid development of MFR, the switch between different work modes is becoming more flexible, increasing the agility of pulse parameters. Most of the existing approaches for recognizing MFR behaviors heavily depend on prior information, which can hardly be obtained in a non-cooperative way. This study develops a novel hierarchical contrastive self-supervise-based method for segmenting and clustering MFR pulse sequences. First, a convolutional neural network (CNN) with a limited receptive field is trained in a contrastive way to distinguish between pulse descriptor words (PDW) in the original order and the samples created by random permutations to detect the boundary between each radar word and perform segmentation. Afterward, the K-means++ algorithm with cosine distances is established to cluster the segmented PDWs according to the output vectors of the CNN’s last layer for radar words extraction. This segmenting and clustering process continues to go in the extracted radar word sequence, radar phase sequence, and so on, finishing the automatic extraction of MFR behavior states in the MFR hierarchical model. Simulation results show that without using any labeled data, the proposed method can effectively mine distinguishable patterns in the sequentially arriving PDWs and recognize the MFR behavior states under corrupted, overlapped pulse parameters.</div>

Attention enhanced capsule network for text classification by encoding syntactic dependency trees with graph convolutional neural network

Text classification is a fundamental task in many applications such as topic labeling, sentiment analysis, and spam detection. The text syntactic relationship and word sequence are important and useful for text classification. How to model and incorporate them to improve performance is one key challenge. Inspired by human behavior in understanding text. In this paper, we combine the syntactic relationship, sequence structure, and semantics for text representation, and propose an attention-enhanced capsule network-based text classification model. Specifically, we use graph convolutional neural networks to encode syntactic dependency trees, build multi-head attention to encode dependencies relationship in text sequence, merge with semantic information by capsule network at last. Extensive experiments on five datasets demonstrate that our approach can effectively improve the performance of text classification compared with state-of-the-art methods. The result also shows capsule network, graph convolutional neural network, and multi-headed attention has integration effects on text classification tasks.

Fast priming of grammatical decisions: repetition and transposed-word priming effects

We used the grammatical decision task to investigate fast priming of written sentence processing. Targets were sequences of 5 words that either formed a grammatically correct sentence or were ungrammatical. Primes were sequences of 5 words and could be the same word sequence as targets, a different sequence of words with a similar syntactic structure, the same sequence with two inner words transposed or the same sequence with two inner words substituted by different words. Prime-word sequences were presented in a larger font size than targets for 200 ms and followed by the target sequence after a 100 ms delay. We found robust repetition priming in grammatical decisions, with same sequence primes leading to faster responses compared with prime sequences containing different words. We also found transposed-word priming effects, with faster responses following a transposed-word prime compared with substituted-word primes. We conclude that fast primed grammatical decisions might offer investigations of written sentence processing what fast primed lexical decisions have offered studies of visual word recognition.

Pseudo-siamese networks with lexicon for Chinese short text matching

Short text matching is one of the fundamental technologies in natural language processing. In previous studies, most of the text matching networks are initially designed for English text. The common approach to applying them to Chinese is segmenting each sentence into words, and then taking these words as input. However, this method often results in word segmentation errors. Chinese short text matching faces the challenges of constructing effective features and understanding the semantic relationship between two sentences. In this work, we propose a novel lexicon-based pseudo-siamese model (CL2 N), which can fully mine the information expressed in Chinese text. Instead of utilizing a character-sequence or a single word-sequence, CL2 N augments the text representation with multi-granularity information in characters and lexicons. Additionally, it integrates sentence-level features through single-sentence features as well as interactive features. Experimental studies on two Chinese text matching datasets show that our model has better performance than the state-of-the-art short text matching models, and the proposed method can solve the error propagation problem of Chinese word segmentation. Particularly, the incorporation of single-sentence features and interactive features allows the network to capture the contextual semantics and co-attentive lexical information, which contributes to our best result.

Rol del tutor en el desarrollo de habilidades cognitivas en alumnos con discapacidad intelectual

La presente investigación tuvo como objetivo diagnosticar las habilidades cognitivas en los alumnos del centro de educación básica especial Unámonos, desde el año 2013 hasta el año 2016. Con tal objetivo se realizó una medición de las habilidades cognitivas de los alumnos del centro el año 2013 (76 estudiantes ) para posteriormente realizar una nueva medición de dichas habilidades el año 2016, luego de la aplicación de 3 años consecutivos del plan de acción individual en cada uno de los alumnos a fin de evaluar los resultados obtenidos, adicionalmente durante los años 2014,2015 se realizó una evaluación (de 23 estudiantes ),para evaluar la eficacia de la estrategia plan de acción individual. Cabe resaltar que para la realización de la presente investigación y como parte del proceso de evaluación de las habilidades cognitivas se procedió a la creación de instrumentos especialmente diseñados y acordes a las necesidades de niños con habilidades diferentes, mismos que fueron creados por la autora, dichos instrumentos incluyen la evaluación del desarrollo integral en alumnos con habilidades diferentes en sus diferentes versiones A, B, C Y D para alumnos cuyas edades oscila entre los 3 y 28 años. La información obtenida en la presente investigación fue recogida a través de evaluaciones individuales con cada uno de los alumnos del centro, tiempo en el cual se tomaron todas las previsiones las cuales incluyen fotografías y videos de los alumnos a fin de alcanzar la mayor validez y confiabilidad del estudio. Cabe resaltar que al finalizar la investigación se observó un incremento significativo en las habilidades cognitivas de cada uno de los alumnos especialmente en áreas tales como: secuencia de palabras, lenguaje comprensivo y nociones espaciales, áreas donde se observaron mejoras en los estudiantes de hasta un 80 % alcanzándose los objetivos generales y específicos propuestos para este trabajo. La hipótesis de la investigación, “todos los alumnos, independientemente del grado de discapacidad que presenten, tienen la posibilidad de mejorar y desarrollar sus habilidades a partir de mediciones adecuadas y de estrategias pedagógicas apropiadas”, fue totalmente comprobada. The objective of this research was to diagnose the cognitive skills of the students of the Unámonos basic special education center, from 2013 to 2016. With such objective, a measurement of the cognitive skills of the students of the center was carried out in 2013 (76 students) to subsequently perform a new measurement of such skills in 2016, after the application of 3 consecutive years of the individual action plan in each of the students in order to evaluate the results obtained, additionally during the years 2014,2015 an evaluation was carried out (of 23 students), to evaluate the effectiveness of the individual action plan strategy. It should be noted that in order to carry out this research and as part of the evaluation process of cognitive skills, we proceeded to the creation of instruments specially designed and according to the needs of children with different abilities, which were created by the author, these instruments include the evaluation of the integral development in students with different abilities in their different versions A, B, C and D for students whose ages range between 3 and 28 years. The information obtained in the present research was collected through individual evaluations with each of the students of the center, during which time all the previsions were taken, including photographs and videos of the students in order to achieve the greatest validity and reliability of the study. It should be noted that at the end of the research a significant increase was observed in the cognitive skills of each of the students, especially in areas such as: word sequence, comprehension language and spatial notions, areas where improvements were observed in the students of up to 80%, reaching the general and specific objectives proposed for this work. The hypothesis of the research, "all students, regardless of the degree of disability they present, have the possibility of improving and developing their skills based on adequate measurements and appropriate pedagogical strategies", was fully proven.

Automated Writing Support for Swedish Learners

This paper describes a tool developed for lexical and grammatical analysis of Swedish text and providing automated feedback for language learners. The system looks for words and word sequences that are likely to contain errors and suggests how to correct them using different non-neural models. The feedback consists of alternative word and word sequence suggestions and morphological features which need to be corrected. Although the system is able to provide reasonable feedback which is believed to be useful for language learners, it still needs further improvements to address the drawbacks such as low precision.

Document Summarization Based on Coverage with Noise Injection and Word Association

Automatic document summarization is a field of natural language processing that is rapidly improving with the development of end-to-end deep learning models. In this paper, we propose a novel summarization model that consists of three methods. The first is a coverage method based on noise injection that makes the attention mechanism select only important words by defining previous context information as noise. This alleviates the problem that the summarization model generates the same word sequence repeatedly. The second is a word association method to update the information of each word by comparing the information of the current step with the information of all previous decoding steps. According to following words, this catches a change in the meaning of the word that has been already decoded. The third is a method using a suppression loss function that explicitly minimizes the probabilities of non-answer words. The proposed summarization model showed good performance on some recall-oriented understudy for gisting evaluation (ROUGE) metrics compared to the state-of-the-art models in the CNN/Daily Mail summarization task, and the results were achieved with very few learning steps compared to the state-of-the-art models.

Reorganization of the neurobiology of language after sentence overlearning

There is a widespread assumption that there are a static set of ‘language regions’ in the brain. Yet, people still regularly produce familiar ‘formulaic’ expressions when those regions are severely damaged. This suggests that the neurobiology of language varies with the extent of word sequence learning and might not be fixed. We test the hypothesis that perceiving sentences is mostly supported by sensorimotor regions involved in speech production and not ‘language regions’ after overlearning. Twelve participants underwent two sessions of behavioural testing and functional magnetic resonance imaging (fMRI), separated by 15 days. During this period, they repeated two sentences 30 times each, twice a day. In both fMRI sessions, participants ‘passively’ listened to those two sentences and novel sentences. Lastly, they spoke novel sentences. Behavioural results confirm that participants overlearned sentences. Correspondingly, there was an increase or recruitment of sensorimotor regions involved in sentence production and a reduction in activity or inactivity for overlearned sentences in regions involved in listening to novel sentences. The global network organization of the brain changed by ∼45%, mostly through lost connectivity. Thus, there was a profound reorganization of the neurobiology of speech perception after overlearning towards sensorimotor regions not considered in most contemporary models and away from the ‘language regions’ posited by those models. These same sensorimotor regions are generally preserved in aphasia and Alzheimer’s disease, perhaps explaining residual abilities with formulaic language. These and other results warrant reconsidering static neurobiological models of language.