Journals

• Question Classification and Answer Extraction for Developing a Cooking QA System
  Published at Computación y Sistemas 2020
  Abdullah Faiz Ur Rahman Khilji, Riyanka Manna, Sahinur Rahman Laskar, Partha Pakray, Dipankar Das, Sivaji, and Gelbukh Bandyopadhyay, Alexander Gelbukh
  PDF Link BIB Abstract
  In an automated Question Answering (QA) system, Question Classification (QC) is an essential module. The aim of QC is to identify the type of questions and classify them based on the expected answer type. Although the machine-learning approach overcomes the limitation of rules as is the case with the conventional rule-based approach but is restricted to the predefined class of questions. The existing approaches are too specific for the users. To address this challenge, we have developed a cooking QA system in which a recipe question is contextually classified into a particular category using deep learning techniques. The question class is then used to extract the requisite details from the recipe obtained via the rule-based approach to provide a precise answer. The main contribution of this paper is the description of the QC module of the cooking QA system. The obtained intermediate classification accuracy over the unseen data is 90% and the human evaluation accuracy of the final system output is 39.33%.

Book Chapter

• Chapter 3: Heal Favor: A Chatbot Application in Healthcare
  Accepted at CRC Press book on Analysis of medical modalities for improved diagnosis in modern healthcare
  Abdullah Faiz Ur Rahman Khilji, Sahinur Rahman Laskar , Partha Pakray, Rabiah Abdul Kadir, Maya Silvi Lydia, Sivaji Bandyopadhyay

Conferences

• EnAsCorp1.0: English-Assamese Corpus
  Accepted at The 3rd Workshop on Technologies for MT of Low Resource Languages (LoResMT 2020), AACL-IJCNLP, December 4–7, 2020
  Sahinur Rahman Laskar, Abdullah Faiz Ur Rahman Khilji, Partha Pakray and Sivaji Bandyopadhyay

• Urdu Fake News Detection using Generalized Autoregressors
  Accepted at Working Notes – The 2020 Fake News Detection in the Urdu Language Task, Forum for Information Retrieval Evaluation 2020, Dec 16-20, 2020, Hyderabad, India
  Abdullah Faiz Ur Rahman Khilji, Sahinur Rahman Laskar, Partha Pakray and Sivaji Bandyopadhyay

• Mathematical Information Retrieval using Formula2Vec Approach
  Accepted at 26th (Virtual) annual International Conference on. Advanced Computing and Communications. (ADCOM 2020), NIT Silchar, Assam, December 16-18, 2020
  Pankaj Dadure, Partha Pakray, and Sivaji Bandyopadhyay

• Multimodal Neural Machine Translation for English to Hindi
  Accepted at Shared Task (Working Notes) – 7th Workshop on Asian Translation (WAT 2020), Hosted by the AACL-IJCNLP 2020
  Sahinur Rahman Laskar, Abdullah Faiz Ur Rahman Khilji, Partha Pakray and Sivaji Bandyopadhyay

• Zero-Shot Neural Machine Translation: Russian-Hindi
  Accepted at Shared Task (Working Notes) – The 3rd Workshop on Technologies for MT of Low Resource Languages (LoResMT 2020), AACL-IJCNLP, December 4–7, 2020
  Sahinur Rahman Laskar, Abdullah Faiz Ur Rahman Khilji, Partha Pakray and Sivaji Bandyopadhyay

• Preliminary Investigation on Causality Information Retrieval
  Accepted at Causality-driven Ad hoc Information Retrieval (CAIR), Forum for Information Retrieval Evaluation 2020, Dec 16-20, 2020, Hyderabad, India
  Pankaj Dadure, Partha Pakray, and Sivaji Bandyopadhyay

• Hindi-Marathi Cross Lingual Model
  Accepted at Shared Task (Working Notes) – (EMNLP 2020) Fifth Conference on Machine Translation (WMT20), November 19-20, 2020
  Sahinur Rahman Laskar, Abdullah Faiz Ur Rahman Khilji, Partha Pakray and Sivaji Bandyopadhyay

• An Analysis of Variable-Size Vector Based Approach for Formula Searching
  Accepted at CLEF 2020 Working Notes Working Notes of CLEF 2020 – Conference and Labs of the Evaluation Forum Thessaloniki, Greece, September 22-25, 2020, Vol-2696, ISSN 1613-0073
  Pankaj Dadure, Partha Pakray, and Sivaji Bandyopadhyay
  PDF BIB Abstract
  In an automated Question Answering (QA) system, Question Classification (QC) is an essential module. The aim of QC is to identify the type of questions and classify them based on the expected answer type. Although the machine-learning approach overcomes the limitation of rules as is the case with the conventional rule-based approach but is restricted to the predefined class of questions. The existing approaches are too specific for the users. To address this challenge, we have developed a cooking QA system in which a recipe question is contextually classified into a particular category using deep learning techniques. The question class is then used to extract the requisite details from the recipe obtained via the rule-based approach to provide a precise answer. The main contribution of this paper is the description of the QC module of the cooking QA system. The obtained intermediate classification accuracy over the unseen data is 90% and the human evaluation accuracy of the final system output is 39.33%.

• HealFavor: Dataset and A Prototype System for Healthcare ChatBot
  Accepted at International Conference on Data Science, Artificial Intelligence and Business Analytics 2020 Medan, North Sumatra, Indonesia
  Abdullah Faiz Ur Rahman Khilji, Sahinur Rahman Laskar , Partha Pakray, Rabiah Abdul Kadir, Maya Silvi Lydia, Sivaji Bandyopadhyay
  BIB Abstract
  A chatbot is a software application aimed at simulating real-time conversations. This system has been designed to address a plethora of domains where they have proved themselves worthy to complement or in some areas replace human-based information acquisition. Though some domains like travel and food have advanced with the growing consumer demand, the healthcare-based system does require significant advancement to address the issue of medical accessibility. The work aims at providing a suitable dataset as well as proposes a prototype system architecture. The prototype system with the self-created dataset is then analyzed on different parameters by numerous experts.

• Neural Machine Translation: Assamese↔Bengali
  Accepted at International Conference on Modeling, Simulation and Optimization (CoMSO 2020) National Institute of Technology Silchar
  Sahinur Rahman Laskar, Partha Pakray and Sivaji Bandyopadhyay

• Neural Machine Translation for Low Resource Assamese-English
  Accepted at 3rd International Conference On Computing and Communication Systems
  Sahinur Rahman Laskar, Partha Pakray and Sivaji Bandyopadhyay

• Efficient Assessment of Formula Representation in Embedded Vector
  Accepted at 3rd International Conference On Computing and Communication Systems
  Pankaj Kundan Dadure, Partha Pakray and Sivaji Bandyopadhyay

• An Empirical Analysis on Retrieval of Math Information from the Scientific Documents
  Accepted at International Conference on Communication and Intelligent System (ICCIS-2019), booktitle
  Pankaj Kundan Dadure, Partha Pakray and Sivaji Bandyopadhyay
  Link BIB Abstract
  Retrieval of math information is one of the recognized domain specific tasks in the research field of Natural Language Processing. Rigorous evaluation in the development of math information retrieval system invented a various indexing and searching techniques. This paper analyzes the contribution of cosine similarity in the retrieval of math information. In the proposed approach, documents are preprocessed by the document and the formula preprocessors, and the preprocessed formula is indexed by the indexer. For retrieving and ranking the documents, the proposed approach uses a cosine similarity measure and based on that the documents are retrieved. The performance of the proposed approach has been tested on the MathTagArticles of Wikipedia of NTCIR-12, and the obtained results show the potential of proposed approach in the retrieval of math information.

• English to Hindi Multi-modal Neural Machine Translation and Hindi Image Captioning
  Accepted at Proceedings of the 6th Workshop on Asian Translation 2019
  Sahinur Rahman Laskar, Rohit Pratap Singh, Partha Pakray, Sivaji Bandyopadhyay
  PDF Link BIB Abstract
  With the widespread use of Machine Trans-lation (MT) techniques, attempt to minimizecommunication gap among people from di-verse linguistic backgrounds. We have par-ticipated in Workshop on Asian Transla-tion 2019 (WAT2019) multi-modal translationtask. There are three types of submissiontrack namely, multi-modal translation, Hindi-only image captioning and text-only transla-tion for English to Hindi translation. The mainchallenge is to provide a precise MT output. The multi-modal concept incorporates textualand visual features in the translation task. Inthis work, multi-modal translation track re-lies on pre-trained convolutional neural net-works (CNN) with Visual Geometry Grouphaving 19 layered (VGG19) to extract imagefeatures and attention-based Neural MachineTranslation (NMT) system for translation. The merge-model of recurrent neural network (RNN) and CNN is used for the Hindi-onlyimage captioning. The text-only translationtrack is based on the transformer model of theNMT system. The official results evaluated atWAT2019 translation task, which shows thatour multi-modal NMT system achieved Bilin-gual Evaluation Understudy (BLEU) score20. 37, Rank-based Intuitive Bilingual Eval-uation Score (RIBES) 0.642838, Adequacy-Fluency Metrics (AMFM) score 0.668260 forchallenge test data and BLEU score 40.55, RIBES 0.760080, AMFM score 0.770860 forevaluation test data in English to Hindi multi-modal translation respectively.

• Neural Machine Translation: English to Hindi
  Accepted at Proceedings of 3rd IEEE Conference on Information & Communication Technology, IIIT Allahabad
  Sahinur Rahman Laskar, Abinash Dutta, Partha Pakray and Sivaji Bandyopadhyay
  PDF Link BIB Abstract
  Machine Translation (MT) attempts to minimize the communication gap among people from various linguistic backgrounds. Automatic translation between pair of different natural languages is the task of MT mechanism, wherein Neural Machine Translation (NMT) attract attention because it offers reasonable translation accuracy in case of the context analysis and fluent translation. In this paper, two different NMT systems are carried out, namely, NMT-1 relies on the Long Short Term Memory (LSTM) based attention model and NMT-2 depends on the transformer model in the context of English to Hindi translation. System results are evaluated using Bilingual Evaluation Understudy (BLEU) metric. The average BLEU scores of NMT-1 system are 35.89 (Test-Set-1), 19.91 (Test-Set-2) and NMT-2 system are 34.42 (Test-Set-1), 24.74 (Test-Set-2) respectively. The results show better performance than existing NMT systems.

• Neural Machine Translation: Hindi-Nepali
  Accepted at Proceedings of the Fourth Conference on Machine Translation (WMT)
  Sahinur Rahman Laskar, Partha Pakray and Sivaji Bandyopadhyay
  PDF Link BIB Abstract
  In an automated With the extensive use of Machine Translation (MT) technology, there is progressively interest in directly translating between pairs of similar languages. Because the main challenge is to overcome the limitation of available parallel data to produce a precise MT output. Current work relies on the Neural Machine Translation (NMT) with attention mechanism for the similar language translation of WMT19 shared task in the context of Hindi-Nepali pair. The NMT systems trained the Hindi-Nepali parallel corpus and tested, analyzed in Hindi⇔ Nepali translation. The official result declared at WMT19 shared task, which shows that our NMT system obtained Bilingual Evaluation Understudy (BLEU) score 24.6 for primary configuration in Nepali to Hindi translation. Also, we have achieved BLEU score 53.7 (Hindi to Nepali) and 49.1 (Nepali to Hindi) in contrastive system type.