Abstract

Learning linguistic generalizations from only a few examples is a challenging task. Recent work has shown that program synthesis – a method to learn rules from data in the form of programs in a domain-specific language – can be used to learn phonological rules in highly data-constrained settings. In this paper, we use the problem of phonological stress placement as a case to study how the design of the domain-specific language influences the generalization ability when using the same learning algorithm. We find that encoding the distinction between consonants and vowels results in much better performance, and providing syllable-level information further improves generalization. Program synthesis, thus, provides a way to investigate how access to explicit linguistic information influences what can be learnt from a small number of examples.

Abstract

In this paper, we present a novel approach for domain adaptation in Neural Machine Translation which aims to improve the translation quality over a new domain. Adapting new domains is a highly challenging task for Neural Machine Translation on limited data, it becomes even more difficult for technical domains such as Chemistry and Artificial Intelligence due to specific terminology, etc. We propose Domain Specific Back Translation method which uses available monolingual data and generates synthetic data in a different way. This approach uses Out Of Domain words. The approach is very generic and can be applied to any language pair for any domain. We conduct our experiments on Chemistry and Artificial Intelligence domains for Hindi and Telugu in both directions. It has been observed that the usage of synthetic data created by the proposed algorithm improves the BLEU scores significantly.

Abstract

This paper describes the participation of team oneNLP (LTRC, IIIT-Hyderabad) for the WMT 2021 task, similar language translation1 . We experimented with transformer based Neural Machine Translation and explored the use of language similarity for Tamil-Telugu and Telugu-Tamil. We incorporated use of different subword configurations, script conversion and single model training for both directions as exploratory experiments.

Abstract

Neural models excel at extracting statistical patterns from large amounts of data, but struggle to learn patterns or reason about language from only a few examples. In this paper, we ask: Can we learn explicit rules that generalize well from only a few examples? We explore this question using program synthesis. We develop a synthesis model to learn phonology rules as programs in a domain-specific language. We test the ability of our models to generalize from few training examples using our new dataset of problems from the Linguistics Olympiad, a challenging set of tasks that require strong linguistic reasoning ability. In addition to being highly sample-efficient, our approach generates human-readable programs, and allows control over the generalizability of the learnt programs.

Abstract

We present findings from a first in-depth postediting effort estimation study in the EnglishHindi direction along multiple effort indicators. We conduct a controlled experiment involving professional translators, who complete assigned tasks alternately, in a translation from scratch and a post-edit condition. We find that post-editing reduces translation time (by 63%), utilizes fewer keystrokes (by 59%), and decreases the number of pauses (by 63%) when compared to translating from scratch. We further verify the quality of translations thus produced via a human evaluation task in which we do not detect any discernible quality differences.

Abstract

This paper describes the work and the systems submitted by the IIIT-Hyderbad team (Id: IIITH) in the WAT 2021 (Nakazawa et al., 2021) MultiIndicMT shared task. The task covers 10 major languages of the Indian subcontinent. For the scope of this task, we have built multilingual systems for 20 translation directions namely English-Indic (one-to-many) and Indic-English (many-to-one). Individually, Indian languages are resource poor which hampers translation quality but by leveraging multilingualism and abundant monolingual corpora, the translation quality can be substantially boosted. But the multilingual systems are highly complex in terms of time as well as computational resources. Therefore, we are training our systems by efficiently selecting data that will actually contribute to most of the learning process. Furthermore, we are also exploiting the language relatedness found in between Indian languages. All the comparisons were made using BLEU score and we found that our final multilingual system significantly outperforms the baselines by an average of 11.3 and 19.6 BLEU points for EnglishIndic (en-xx) and Indic-English (xx-en) directions, respectively.

Abstract

India is one of the most linguistically diverse nations of the world and is culturally very rich. Most of these languages are somewhat similar to each other on account of sharing a common ancestry or being in contact for a long period of time (Bhattacharyya et al., 2016). Nowadays, researchers are constantly putting efforts in utilizing the language relatedness to improve the performance of various NLP systems such as cross lingual semantic search, machine translation (Kunchukuttan and Bhattacharyya, 2020), sentiment analysis systems, etc. So in this paper, we performed an extensive case study on similarity involving languages of the Indian subcontinent. Language similarity prediction is defined as the task of measuring how similar the two languages are on the basis of their lexical, morphological and syntactic features. In this study, we concentrate only on the approach to calculate lexical similarity between Indian languages by looking at various factors such as size and type of corpus, similarity algorithms, subword segmentation, etc. The main takeaways from our work are: (i) Relative order of the language similarities largely remain the same, regardless of the factors mentioned above, (ii) Similarity within the same language family is higher, (iii) Languages share more lexical features at the subword level.

Abstract

India is known as the land of many tongues and dialects. Neural machine translation (NMT) is the current state-of-the-art approach for machine translation (MT) but performs better only with large datasets which Indian languages usually lack, making this approach infeasible. So, in this paper, we address the problem of data scarcity by efficiently training multilingual and multilingual multi domain NMT systems involving languages of the Indian subcontinent. We are proposing the technique for using the joint domain and language tags in a multilingual setup. We draw three major conclusions from our experiments: (i) Training a multilingual system via exploiting lexical similarity based on language family helps in achieving an overall average improvement of 3.25 BLEU points over bilingual baselines, (ii) Technique of incorporating domain information into the language tokens helps multilingual multi-domain system in getting a significant average improvement of 6 BLEU points over the baselines, (iii) Multistage fine-tuning further helps in getting an improvement of 1- 1.5 BLEU points for the language pair of interest.

Abstract

India is known as the land of many tongues. There is no single language called Indian. India speaks hundreds of languages and dialects. Some are extinct, while some are still in use with considerable speakers. Despite having a lot of different scripts, most of the Indian languages still share a lot of lexical features which can be utilized to help improve the quality of Multilingual NMT systems trained on them. So, in this paper, we present an extensive study of Multilingual as well as Multilingual Multi Domain NMT involving languages of the Indian subcontinent. We draw four major conclusions from our experiments: (i) Multilingual Multi Domain models can significantly improve the accuracy of all the individual languages within their domains, resulting in improving the overall performance of the Multilingual Multi Domain system, (ii) Encoder representation of different languages based on their family helps Multilingual models gain an average improvement of 3.25 BLEU points, (iii) Our new technique of incorporating domain information into the language tokens results in getting a significant improvement of 6 BLEU points on an average as compared to the baselines, (iv) Multistage Fine-tuning further helps in improvement of (1-1.5) BLEU points.

Abstract

The introduction of transformers in 2017 and successively BERT in 2018 brought about a revolution in the field of natural language processing. Such models are pretrained on vast amounts of data, and are easily extensible to be used for a wide variety of tasks through transfer learning. Continual work on transformer based architectures has led to a variety of new models with state of the art results. RoBERTa (Liu et al., 2019) is one such model, which brings about a series of changes to the BERT (Devlin et al., 2018) architecture and is capable of producing better quality embeddings at an expense of functionality. In this paper, we attempt to solve the well known text classification task of fine-grained domain classification using BERT and RoBERTa and perform a comparative analysis of the same. We also attempt to evaluate the impact of data preprocessing specially in the context of fine-grained domain classification. The results obtained outperformed all the other models at the ICON TechDOfication 2020 (subtask-2a) Fine-grained domain classification task and ranked first. This proves the effectiveness of our approach.