@nitmz.ac.in
PGT Computer Science
Kendriya Vidyalaya Khairagarh
PhD NIT Mizoram
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Amarnath Pathak and Partha Pakray
SAGE Publications
The article presents an approach to recognise formula entailment, which concerns finding entailment relationships between pairs of math formulae. As the current formula-similarity-detection approaches fail to account for broader relationships between pairs of math formulae, recognising formula entailment becomes paramount. To this end, a long short-term memory (LSTM) neural network using symbol-by-symbol attention for recognising formula entailment is implemented. However, owing to the unavailability of relevant training and validation corpora, the first and foremost step is to create a sufficiently large-sized symbol-level MATHENTAIL data set in an automated fashion. Depending on the extent of similarity between the corresponding symbol embeddings, the symbol pairs in the MATHENTAIL data set are assigned ‘entailment’ or ‘neutral’ labels. An improved symbol-to-vector (isymbol2vec) method generates mathematical symbols (in LATEX) and their embeddings using the Wikipedia corpus of scientific documents and Continuous Bag of Words (CBOW) architecture. Eventually, the LSTM network, trained and validated using the MATHENTAIL data set, predicts formulae entailment for test formulae pairs with a reasonable accuracy of 62.2%.
Amarnath Pathak, Partha Pakray, and Ranjita Das
Springer Nature Singapore
Amarnath Pathak, Riyanka Manna, Partha Pakray, Dipankar Das, Alexander Gelbukh, and Sivaji Bandyopadhyay
Springer Science and Business Media LLC
Amarnath Pathak, Partha Pakray, and Jereemi Bentham
Springer Science and Business Media LLC
Amarnath Pathak and Partha Pakray
Walter de Gruyter GmbH
Abstract Machine Translation bridges communication barriers and eases interaction among people having different linguistic backgrounds. Machine Translation mechanisms exploit a range of techniques and linguistic resources for translation prediction. Neural machine translation (NMT), in particular, seeks optimality in translation through training of neural network, using a parallel corpus having a considerable number of instances in the form of a parallel running source and target sentences. Easy availability of parallel corpora for major Indian language forms and the ability of NMT systems to better analyze context and produce fluent translation make NMT a prominent choice for the translation of Indian languages. We have trained, tested, and analyzed NMT systems for English to Tamil, English to Hindi, and English to Punjabi translations. Predicted translations have been evaluated using Bilingual Evaluation Understudy and by human evaluators to assess the quality of translation in terms of its adequacy, fluency, and correspondence with human-predicted translation.
Amarnath Pathak, Partha Pakray, and Ranjita Das
IEEE
The work presented in this paper ascertains role of Long Sort-Term Memory (LSTM) neural network in Math Information Retrieval (MIR). Motivated from promising performances of the LSTM for sequence-to-sequence tasks, an LSTM based Formula Entailment (LFE) module is implemented for recognizing entailment between mathematical user query and document formulae. The LFE module is trained and validated using a symbol level Math Formula Entailment (MENTAIL) dataset. The relevance of a document is determined by the fraction of document formulae which entail the user query. A reasonable score of 0.45 for the P_5 evaluation measure substantiates competence of the implemented MIR system in retrieving relevant documents corresponding to a mathematical user query.
Amarnath Pathak, Ranjita Das, Partha Pakray, and Alexander Gelbukh
Instituto Politecnico Nacional/Centro de Investigacion en Computacion
A math formula present inside a scientific document is often preceded by its textual description, which is commonly referred to as the context of formula. Annotating context to the formula enriches its semantics, and consequently impacts the retrieval of mathematical contents from scientific documents. Also, with a considerable surety, a context can be assumed to be one of the Noun Phrases (NPs) of the sentence in which formula occurs. However, the presence of several different misleading NPs in the sentence necessitates extraction of an NP, which is more precise to the formula than the rest. Although a fair number of methods are developed for precise context extraction, it can be fascinating to prospect other competent techniques which can further their performances. To this end, this paper discusses implementation of an automated context extraction system, which follows certain heuristics in assigning weights to different candidate NPs, and tune those weights using a development set comprising annotated formulae. The implemented system significantly outperforms nearest noun and sentence–pattern based methods on the ground of F–score.
Amarnath Pathak, Partha Pakray, and Alexander Gelbukh
IOS Press
Amarnath Pathak, Dhruv Goel, and Somen Debnath
Springer Singapore
Amarnath Pathak, Partha Pakray, and Alexander Gelbukh
Instituto Politecnico Nacional/Centro de Investigacion en Computacion
Intricate math formulae, which majorly constitute the content of scientific documents, add to the complexity of scientific document retrieval. Although modifications in conventional indexing and search mechanisms have eased the complexity and exhibited notable performance, the formula embedding approach to scientific document retrieval sounds equally appealing and promising. Formula Embedding Module of the proposed system uses a Bit Position Information Table to transform math formulae, contained inside scientific documents, into binary formulae vectors. Each set bit of a formula vector designates presence of aspecific mathematical entity. Mathematical user query is transformed into query vector, in similar fashion, and the corresponding relevant documents are retrieved. Relevance of a search result is characterized by extent of similarity between the indexed formula vector and the query vector. Promising performance, under moderately constrained situation, substantiates competence of the proposed approach.
Partha Pakray, Goutam Majumder, and Amarnath Pathak
Springer Singapore
Amarnath Pathak and Partha Pakray
Springer Singapore
Amarnath Pathak, Partha Pakray, Sandip Sarkar, Dipankar Das, and Alexander Gelbukh
Instituto Politecnico Nacional/Centro de Investigacion en Computacion
Effective retrieval of mathematical contents from vast corpus of scientific documents demands enhancement in the conventional indexing and searching mechanisms. Indexing mechanism and the choice of semantic similarity measures guide the results of Math Information Retrieval system (MathIRs) to perfection. Tokenization and formula unification are among the distinguishing i features of indexing mechanism, used in MathIRs, which facilitate sub-formula and similarity search. Besides, the scientific documents and the user queries in MathIRs will contain math as well as text contents and to match these contents we require three important modules: Text-Text Similarity (TS), Math-Math Similarity (MS) and Text-Math Similarity (TMS). In this paper we have proposed MathIRs comprising these important modules and a substitution tree based mechanism for indexing mathematical expressions. We have also presented experimental results for similarity search and argued that proposal of MathIRs will ease the task of scientific document retrieval.