Terms Of BLEU Score Research Articles

TransCP: A Transformer Pointer Network for Generic Entity Description Generation With Explicit Content-Planning

We study neural data-to-text generation to generate a sentence to describe a target entity based on its attributes. Specifically, we address two problems of the encoder-decoder framework for data-to-text generation: i) how to encode a non-linear input (e.g., a set of attributes); and ii) how to order the attributes in the generated description. Existing studies focus on the encoding problem but do not address the ordering problem, i.e., they learn the content-planning implicitly. The other approaches focus on two-stage models but overlook the encoding problem. To address the two problems at once, we propose a model named <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">TransCP</b> to explicitly learn content-planning and integrate them into a description generation model in an end-to-end fashion. We propose a novel Transformer-based Pointer Network with <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">gated residual attention</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">importance masking</i> to learn a content-plan. To integrate the content-plan with a description generator, we propose a tracking mechanism to trace the extent to which the content-plan is exposed in the previous decoding time-step. This helps the description generator select the attributes to be mentioned in proper order. Experimental results show that our model consistently outperforms state-of-the-art baselines by up to <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$2\%$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$3\%$</tex-math></inline-formula> in terms of BLEU score on two real-world datasets.

GAGPT-2: A Geometric Attention-based GPT-2 Framework for Image Captioning in Hindi

Image captioning frameworks usually employ an encoder-decoder paradigm, with the encoder receiving abstract image feature vectors as input and the decoder for language modeling. Nowadays, most prominent architectures employ features from region proposals derived from object detection modules. In this work, we propose a novel architecture for image captioning. We employ the object detection module integrated with transformer architecture as an encoder and GPT-2 (Generative Pre-trained Transformer) as a decoder. The encoder utilizes the information of the spatial relationships among detected objects. We introduce a unique methodology for image caption generation in Hindi, which is widely spoken in South Asia and India and is the world’s third most spoken language as well as India’s official language. In terms of BLEU scores, the proposed approach’s performance is comparable to those of other baselines, and the results illustrate that the proposed approach outperforms the other baselines. The efficacy of the proposed approach in generating correct captions is further determined by human assessment in terms of adequacy and fluency.

Dynamic Convolution-based Encoder-Decoder Framework for Image Captioning in Hindi

In sequence-to-sequence modeling tasks, such as image captioning, machine translation, and visual question answering, encoder-decoder architectures are state of the art. An encoder, convolutional neural network (CNN) encodes input images into fixed dimensional vector representation in the image captioning task, whereas a decoder, a recurrent neural network, performs language modeling and generates the target descriptions. Recent CNNs use the same operation over every pixel; however, all the image pixels are not equally important. To address this, the proposed method uses a dynamic convolution-based encoder for image encoding or feature extraction, Long-Short-Term-Memory as a decoder for language modeling, and X-Linear attention to make the system robust. Encoders, attentions, and decoders are important aspects of the image captioning task; therefore, we experiment with various encoders, decoders, and attention mechanisms. Most of the works for image captioning have been carried out for the English language in the existing literature. We propose a novel approach for caption generation from images in Hindi. Hindi, widely spoken in South Asia and India, is the fourth most-spoken language globally; it is India’s official language. The proposed method utilizes dynamic convolution operation on the encoder side to obtain a better image encoding quality. The Hindi image captioning dataset is manually created by translating the popular MSCOCO dataset from English to Hindi. In terms of BLEU scores, the performance of the proposed method is compared with other baselines, and the results obtained show that the proposed method outperforms different baselines. Manual human assessment in terms of adequacy and fluency of the captions generated further determines the efficacy of the proposed method in generating good-quality captions.

Hybrid Architecture using CNN and LSTM for Image Captioning in Hindi Language

With the advent of deep learning in recent years, the integration of computer vision with natural language processing has garnered a lot of attention. Generating descriptions from images is one of the most intriguing and focused areas of Machine Learning that faces a number of obstacles, particularly when describing images in languages other than English. In image captioning, a computer is trained to understand the visual information of an image and then generate a description based on the image features and reference sentences. Having an application that automatically describes events in their environment and then translates them into a caption or message can make a significant contribution to society. This paper presents a multi-layered CNN-LSTM neural network model that is utilized to recognize and generate Hindi captions for the objects in images. In addition, a variety of models were trained by adjusting hyperparameters and the number of hidden layers to find the optimum model and maximize the likelihood of the resultant Hindi description. Moreover, after testing the effectiveness of our models, it was observed that our model has shown an increase of 34.64% and 29.13% in terms of BLEU score (Unigram) and BLEU score (Bigram) respectively when compared to the existing work done in this field. Image captioning in Hindi can have a multitude of applications in today's society, and it can also provide a user-friendly interface for Hindi speakers.

One-to-Many Semantic Communication Systems: Design, Implementation, Performance Evaluation

Semantic communication in the 6G era has been deemed a promising communication paradigm to break through the bottleneck of traditional communications. However, its applications for the multi-user scenario, especially the broadcasting case, remain under-explored. To effectively exploit the benefits enabled by semantic communication, in this paper, we propose a one-to-many semantic communication system. Specifically, we propose a deep neural network (DNN) enabled semantic communication system called MR_DeepSC. By leveraging semantic features for different users, a semantic recognizer based on the pre-trained model, i.e., DistilBERT, is built to distinguish different users. Furthermore, the transfer learning is adopted to speed up the training of new receiver networks. Simulation results demonstrate that the proposed MR_DeepSC can achieve the best performance in terms of BLEU score than the other benchmarks under different channel conditions, especially in the low signal-to-noise ratio (SNR) regime.

Evaluating the English-Turkish parallel treebank for machine translation

This study extends our initial efforts in building an English-Turkish parallel treebank corpus for statistical machine translation tasks. We manually generated parallel trees for about 17K sentences selected from the Penn Treebank corpus. English sentences vary in length: 15 to 50 tokens including punctuation. We constrained the translation of trees by (i) reordering of leaf nodes based on suffixation rules in Turkish, and (ii) gloss replacement. We aim to mimic human annotator?s behavior in real translation task. In order to fill the morphological and syntactic gap between languages, we do morphological annotation and disambiguation. We also apply our heuristics by creating Nokia English-Turkish Treebank (NTB) to address technical document translation tasks. NTB also includes 8.3K sentences in varying lengths. We validate the corpus both extrinsically and intrinsically, and report our evaluation results regarding perplexity analysis and translation task results. Results prove that our heuristics yield promising results in terms of perplexity and are suitable for translation tasks in terms of BLEU scores.

Captioning model based on meta-learning using prior-convergence knowledge for explainable images

Big data has a variety of data types, including image and text. In particular, image data-based research on face recognition and objection detection has been conducted in diverse areas. Deep learning needs a massive amount of data for learning a model accurately. The amount of data collected is different in each area, and thus it is likely to lack data for analysis through deep learning. Accordingly, it is necessary a method of learning a model effectively and predicting a result accurately with the use of a small amount of data. Also, captions and tags are generated to obtain image information. In the case of tagging, an image is expressed with words, while, in the case of captioning, a sentence can be created in connection with words. For this reason, it is possible to obtain image information in detail through captioning, compared to tagging. However, when a caption is created with words, there is the limitation of end-to-end to lower performance if labeled data are not sufficient. As a solution to the problem, meta-learning, in which a small amount of data can be used, is applied. This study proposes the captioning model based on meta-learning using prior-convergence knowledge for explainable images. The proposed method collects multimodal image data for predicting image information. From the collected data, the attributes representing object information and context information are used. After that, with the use of a small amount of data, meta-learning is applied in a bottom-up approach to creating a sentence for captioning. It can solve the problem caused by data shortage. Lastly, for the extraction of image features, LSTM for convolution network and captioning is established, and the basis for explanation is generated through the reverse operation. The generated basis is an image object. An appropriate explanation sentence is displayed in line with a particular object. Performance evaluation is conducted in two ways for accuracy. Firstly, BLEU score is evaluated according to whether there is meta-learning. Secondly, the proposed captioning model based on prior knowledge, RNN-based captioning model, and bidirectional RNN-based captioning model is evaluated in terms of BLEU score. Therefore, through the proposed method, LSTM’s bottom-up method reduces the cost of improving image resolution and solves the data shortage problem through meta-learning. In addition, it is possible to find the basis of image information using subtitles and to more accurately describe information about photos based on XAI.

A Hindi Image Caption Generation Framework Using Deep Learning

Image captioning is the process of generating a textual description of an image that aims to describe the salient parts of the given image. It is an important problem, as it involves computer vision and natural language processing, where computer vision is used for understanding images, and natural language processing is used for language modeling. A lot of works have been done for image captioning for the English language. In this article, we have developed a model for image captioning in the Hindi language. Hindi is the official language of India, and it is the fourth most spoken language in the world, spoken in India and South Asia. To the best of our knowledge, this is the first attempt to generate image captions in the Hindi language. A dataset is manually created by translating well known MSCOCO dataset from English to Hindi. Finally, different types of attention-based architectures are developed for image captioning in the Hindi language. These attention mechanisms are new for the Hindi language, as those have never been used for the Hindi language. The obtained results of the proposed model are compared with several baselines in terms of BLEU scores, and the results show that our model performs better than others. Manual evaluation of the obtained captions in terms of adequacy and fluency also reveals the effectiveness of our proposed approach. Availability of resources : The codes of the article are available at https://github.com/santosh1821cs03/Image_Captioning_Hindi_Language ; The dataset will be made available: http://www.iitp.ac.in/∼ai-nlp-ml/resources.html .

Normalisation of Indonesian-English Code-Mixed Text and its Effect on Emotion Classification

Usage of code-mixed text has increased in re-cent years among Indonesian internet users, who often mix Indonesian-language with English-language text. Normalisation of this code-mixed text into Indonesian needs to be performed to capture the meaning of English parts of the text and process them effectively. We improve a state-of-the-art code-mixed Indonesian-English normalisation system by modifying its pipeline modules. We further analyse the effect of code-mixed normalisation on emotion classification tasks. Our approach significantly improved on a state-of-the-art Indonesian-English code-mixed text normal-isation system in both the individual pipeline modules and the overall system. The new feature set in the language identification module showed an improvement of 4.26% in terms of F1 score. The combination of machine translation and ruleset in the lexical normalisation module improved BLEU score by 25.22% and lowered WER by 62.49%. The use of context in the translation module improved BLEU score by 2.5% and lowered WER by 8.84%. The effectiveness of the overall pipeline normalisation system increased by 32.11% and 33.82%, in terms of BLEU score and WER, respectively. Code-mixed normalisation also improved the accuracy of emotion classification by up to 37.74% in terms of F1 score.

Joint Learning of Question Answering and Question Generation

Question answering (QA) and question generation (QG) are closely related tasks that could improve each other; however, the connection of these two tasks is not well explored in the literature. In this paper, we present two training algorithms for learning better QA and QG models through leveraging one another. The first algorithm extends Generative Adversarial Network (GAN), which selectively incorporates artificially generated instances as additional QA training data. The second algorithm is an extension of dual learning, which incorporates the probabilistic correlation of QA and QG as additional regularization in training objectives. To test the scalability of our algorithms, we conduct experiments on both document based and table based question answering tasks. Results show that both algorithms improve a QA model in terms of accuracy and QG model in terms of BLEU score. Moreover, we find that the performance of a QG model could be easily improved by a QA model via policy gradient, however, directly applying GAN that regards all the generated questions as negative instances could not improve the accuracy of the QA model. Our algorithm that selectively assigns labels to generated questions would bring a performance boost.

Sentence Generation for Entity Description with Content-Plan Attention

We study neural data-to-text generation. Specifically, we consider a target entity that is associated with a set of attributes. We aim to generate a sentence to describe the target entity. Previous studies use encoder-decoder frameworks where the encoder treats the input as a linear sequence and uses LSTM to encode the sequence. However, linearizing a set of attributes may not yield the proper order of the attributes, and hence leads the encoder to produce an improper context to generate a description. To handle disordered input, recent studies propose two-stage neural models that use pointer networks to generate a content-plan (i.e., content-planner) and use the content-plan as input for an encoder-decoder model (i.e., text generator). However, in two-stage models, the content-planner may yield an incomplete content-plan, due to missing one or more salient attributes in the generated content-plan. This will in turn cause the text generator to generate an incomplete description. To address these problems, we propose a novel attention model that exploits content-plan to highlight salient attributes in a proper order. The challenge of integrating a content-plan in the attention model of an encoder-decoder framework is to align the content-plan and the generated description. We handle this problem by devising a coverage mechanism to track the extent to which the content-plan is exposed in the previous decoding time-step, and hence it helps our proposed attention model select the attributes to be mentioned in the description in a proper order. Experimental results show that our model outperforms state-of-the-art baselines by up to 3% and 5% in terms of BLEU score on two real-world datasets, respectively.

Annotated Guidelines and Building Reference Corpus for Myanmar-English Word Alignment

Reference corpus for word alignment is an important resource for developing and evaluating word alignment methods. For Myanmar-English language pairs, there is no reference corpus to evaluate the word alignment tasks. Therefore, we created the guidelines for Myanmar-English word alignment annotation between two languages over contrastive learning and built the Myanmar-English reference corpus consisting of verified alignments from Myanmar ALT of the Asian Language Treebank (ALT). This reference corpus contains confident labels sure (S) and possible (P) for word alignments which are used to test for the purpose of evaluation of the word alignments tasks. We discuss the most linking ambiguities to define consistent and systematic instructions to align manual words. We evaluated the results of annotators agreement using our reference corpus in terms of alignment error rate (AER) in word alignment tasks and discuss the words relationships in terms of BLEU scores.

Graph Based Translation Memory for Neural Machine Translation

A translation memory (TM) is proved to be helpful to improve neural machine translation (NMT). Existing approaches either pursue the decoding efficiency by merely accessing local information in a TM or encode the global information in a TM yet sacrificing efficiency due to redundancy. We propose an efficient approach to making use of the global information in a TM. The key idea is to pack a redundant TM into a compact graph and perform additional attention mechanisms over the packed graph for integrating the TM representation into the decoding network. We implement the model by extending the state-of-the-art NMT, Transformer. Extensive experiments on three language pairs show that the proposed approach is efficient in terms of running time and space occupation, and particularly it outperforms multiple strong baselines in terms of BLEU scores.

Scheduled Multi-Task Learning: From Syntax to Translation

Neural encoder-decoder models of machine translation have achieved impressive results, while learning linguistic knowledge of both the source and target languages in an implicit end-to-end manner. We propose a framework in which our model begins learning syntax and translation interleaved, gradually putting more focus on translation. Using this approach, we achieve considerable improvements in terms of BLEU score on relatively large parallel corpus (WMT14 English to German) and a low-resource (WIT German to English) setup.

Efficient Phrase Table pruning for Hindi to English machine translation through syntactic and marker-based filtering and hybrid similarity measurement

AbstractFor a given training corpus of parallel sentences, the quality of the output produced by a translation system relies heavily on the underlying similarity measurement criteria. A phrase-based machine translation system derives its output through a generative process using a Phrase Table comprising source and target language phrases. As a consequence, the more effective the Phrase Table is, in terms of its size and the output that may be derived out of it, the better is the expected outcome of the underlying translation system. However, finding the most similar phrase(s) from a given training corpus that can help generate a good quality translation poses a serious challenge. In practice, often there are many parallel phrase entries in a Phrase Table that are either redundant, or do not contribute to the translation results effectively. Identifying these candidate entries and removing them from the Phrase Table will not only reduce the size of the Phrase Table, but should also help in improving the processing speed for generating the translations. The present paper develops a scheme based on syntactic structure and the marker hypothesis (Green 1979, The necessity of syntax markers: two experiments with artificial languages, Journal of Verbal Learning and Behavior) for reducing the size of a Phrase Table, without compromising much on the translation quality of the output, by retaining the non-redundant and meaningful parallel phrases only. The proposed scheme is complemented with an appropriate similarity measurement scheme to achieve maximum efficiency in terms of BLEU scores. Although designed for Hindi to English machine translation, the overall approach is quite general, and is expected to be easily adaptable for other language pairs as well.

Factored Statistical Machine Translation for German-English

Machine Translation is a machine that is going to automatically translate given sentences in a language to other particular language. This paper aims to test the effectiveness of a new model of machine translation which is factored machine translation. We compare the performance of the unfactored system as our baseline compared to the factored model in terms of BLEU score. We test the model in German-English language pair using Europarl corpus. The tools we are using is called MOSES. It is freely downloadable and use. We found, however, that the unfactored model scored over 24 in BLEU and outperforms the factored model which scored below 24 in BLEU for all cases. In terms of words being translated, however, all of factored models outperforms the unfactored model.

Fine-grained attention mechanism for neural machine translation

Abstract Neural machine translation (NMT) has been a new paradigm in machine translation, and the attention mechanism has become the dominant approach with the state-of-the-art records in many language pairs. While there are variants of the attention mechanism, all of them use only temporal attention where one scalar value is assigned to one context vector corresponding to a source word. In this paper, we propose a fine-grained (or 2D) attention mechanism where each dimension of a context vector will receive a separate attention score. In experiments with the task of En-De and En-Fi translation, the fine-grained attention method improves the translation quality in terms of BLEU score. In addition, our alignment analysis reveals how the fine-grained attention mechanism exploits the internal structure of context vectors.

Hybrid Arabic–French machine translation using syntactic re-ordering and morphological pre-processing

Arabic is a highly inflected language and a morpho-syntactically complex language with many differences compared to several languages that are heavily studied. It may thus require good pre-processing as it presents significant challenges for Natural Language Processing (NLP), specifically for Machine Translation (MT). This paper aims to examine how Statistical Machine Translation (SMT) can be improved using rule-based pre-processing and language analysis. We describe a hybrid translation approach coupling an Arabic–French statistical machine translation system using the Moses decoder with additional morphological rules that reduce the morphology of the source language (Arabic) to a level that makes it closer to that of the target language (French). Moreover, we introduce additional swapping rules for a structural matching between the source language and the target language. Two structural changes involving the positions of the pronouns and verbs in both the source and target languages have been attempted. The results show an improvement in the quality of translation and a gain in terms of BLEU score after introducing a pre-processing scheme for Arabic and applying these rules based on morphological variations and verb re-ordering (VS into SV constructions) in the source language (Arabic) according to their positions in the target language (French). Furthermore, a learning curve shows the improvement in terms on BLEU score under scarce- and large-resources conditions. The proposed approach is completed without increasing the amount of training data or radically changing the algorithms that can affect the translation or training engines.

Distance-based Reordering in English to Hindi Statistical Machine Translation

paper compares different reordering models on English to Hindi statistical machine translation system. The two Indo- European languages differ significantly in their word order preferences. While English follows SVO model, Hindi follows SOV model. Therefore both long distance and short distance reordering becomes important. The reordering models available in MOSES SMT are discussed and compared with a more novel approach called distance-based reordering. This new approach significantly improves the quality of English to Hindi translation, both in terms of BLEU score and subjective human evaluation..

Parallel Phrase Scoring for Extra-large Corpora

This paper presents a C++ implementation of the phrase scoring step in phrase-based systems that helps to exploit the available computing resources more efficiently and trains very large systems in reasonable time without sacrificing the system’s performance in terms of Bleu score. Three parallelizing tools are made freely available. The first exploits shared memory parallelism and multiple disks for parallel IOs while the two others run in a distributed environment. We demonstrate the efficiency and consistency of our tools, in the framework of the Fr-En systems we developed for the WMT and IWSLT evaluation campaigns, in which we were able to generate the phrase table in one third up to one seventh of the time taken by Moses in the same tasks.