Abstract

Product reviews on e-commerce platforms are a critical form of user-generated content that influence consumer decisions. However, these reviews are predominantly in English, creating a significant accessibility barrier for users who are not fluent in English. When translating into major Indian languages using the current models, the outputs often fail to capture domain-specific features and colloquial style, resulting in stylistically unnatural texts. To address this gap, we introduce textbf{STAR-IL}, a human-annotated, multilingual, parallel corpus for style-aware translation of product reviews. We evaluate the performance of several state-of-the-art models on our dataset for the task of product review translation. Our experiments show that models fine-tuned on STAR-IL achieve significant average performance gain of textbf{5.77} points in BLEU and textbf{3.78} points in COMET, when compared to their baselines, across all languages. Our dataset provides a valuable benchmark for future research in style-aware product review translation. The STAR-IL dataset is publicly available at https://github.com/ltrc/STAR-IL-Corpus

Abstract

Large multilingual transformers such as XLM-RoBERTa achieve impressive performance on diverse NLP benchmarks, but understanding how they internally encode grammatical information remains challenging. This study investigates the encoding of syntactic and morphological information derived from the Paninian grammatical framework--specifically designed for morphologically rich Indian languages--across model layers. Using diagnostic probing, we analyze the hidden representations of frozen XLM-RoBERTa-base, mBERT, and IndicBERT models across seven Indian languages (Hindi, Kannada, Malayalam, Marathi, Telugu, Urdu, Bengali). Probes are trained to predict Paninian dependency relations (by edge probing) and essential morphosyntactic features (UPOS tags, Vibhakti markers). We find that syntactic structure (dependencies) is primarily encoded in the middle-to-upper-middle layers (layers 6-9), while lexical features peak slightly earlier. Although the general layer-wise trends are shared across models, significant variations in absolute probing performance reflect differences in model capacity, pre-training data, and language-specific characteristics. These findings shed light on how theory-specific grammatical information emerges implicitly within multilingual transformer representations trained largely on unstructured raw text.

Abstract

The relationship between tokenizer algorithm (e.g., Byte-Pair Encoding (BPE), Unigram), morphological alignment, tokenization quality (e.g., compression efficiency), and downstream performance remains largely unclear, particularly for languages with complex morphology. In this paper, we conduct a comprehensive evaluation of tokenizers using small-sized BERT models -- from pre-training through fine-tuning -- for Telugu (agglutinative), along with preliminary evaluation in Hindi (primarily fusional with some agglutination) and English (fusional). To evaluate morphological alignment of tokenizers in Telugu, we create a dataset containing gold morpheme segmentations of 600 derivational and 7000 inflectional word forms.
Our experiments reveal two key findings for Telugu. First, the choice of tokenizer algorithm is the most significant factor influencing performance, with Unigram-based tokenizers consistently outperforming BPE across most settings. Second, while better morphological alignment shows a moderate, positive correlation with performance on text classification and structure prediction tasks, its impact is secondary to the tokenizer algorithm. Notably, hybrid approaches that use morphological information for pre-segmentation significantly boost the performance of BPE, though not Unigram. Our results further showcase the need for comprehensive intrinsic evaluation metrics for tokenizers that could explain downstream performance trends consistently.

Abstract

This study addresses the critical challenge of
data scarcity in machine translation for Indian
languages, particularly given their morpholog-
ical complexity and limited parallel data. We
investigate an effective strategy to maximize
the utility of existing data by generating nega-
tive samples from positive training instances us-
ing a progressive perturbation approach. This
is used to align the model with preferential
data using Kahneman-Tversky Optimization
(KTO). Comparing it against traditional Su-
pervised Fine-Tuning (SFT), we demonstrate
how generating negative samples and leverag-
ing KTO enhances data efficiency. By creat-
ing rejected samples through progressively per-
turbed translations from the available dataset,
we fine-tune the Llama 3.1 Instruct 8B model
using QLoRA across 16 language directions, in-
cluding English, Hindi, Bangla, Tamil, Telugu,
and Santali. Our results show that KTO-based
preference alignment with progressive pertur-
bation consistently outperforms SFT, achieving
significant gains in translation quality with an
average BLEU increase of 1.84 to 2.47 and
CHRF increase of 2.85 to 4.01 compared to
SFT for selected languages, while using the
same positive training samples and under simi-
lar computational constraints. This highlights
the potential of our negative sample genera-
tion strategy within KTO, especially in low-
resource scenarios.

Abstract

With the primary focus on evaluating the ef- fectiveness of large language models for au- tomatic reference-less translation assessment, this work presents our experiments on mimick- ing human direct assessment to evaluate the quality of translations in English and Indian languages. We constructed a translation evalua- tion task where we performed zero-shot learn- ing, in-context example-driven learning, and fine-tuning of large language models to pro- vide a score out of 100, where 100 represents a perfect translation and 1 represents a poor translation. We compared the performance of our trained systems with existing methods such as COMET, BERT-Scorer, and LABSE, and found that the LLM-based evaluator (LLaMA- 2-13B) achieves a comparable or higher overall correlation with human judgments for the con- sidered Indian language pairs (Refer figure 1).

Abstract

The objective of the MTIL track in FIRE 2023 was to encourage the development of Indian Language to Indian Language (IL-IL) Neural Machine Translation models. The languages covered in the track included Hindi, Gujarati, Kannada, Odia, Punjabi, Urdu, Telugu, Kashmiri, and Sindhi. The track consists of two tasks: (i) a General Translation Task and (ii) a Domain specific Translation Task with Governance and Healthcare being the chosen domains. For the listed languages, we proposed 12 diverse language directions for the general domain translation task and 8 each for healthcare and governance domains. Participants were encouraged to submit models for one or more language pairs. We witnessed the creation of 34 distinct models spanning various language pairs and domains. Model assessments were conducted using five evaluation metrics: BLEU, CHRF, CHRF++, TER, and COMET. The submitted model outputs were ranked based on the CHRF score.

Abstract

How can we effectively estimate the quality of translated texts in the medical field, where back-translation is
usually available and/or recommended for sensitive documents. This paper proposes a novel metric, GATE1
,
for translation quality estimation task, leveraging the Resource Description Framework (RDF) to encode both
semantic and syntactical information of the original and back-translated sentences into RDF graphs. The distance
between these graphs is measured to get the semantic similarity score to assess the quality of the translation.
Unlike traditional metrics like BLEU and METEOR, our approach is reference-less, capturing both semantic and
syntactical information for a comprehensive assessment of translation quality. Our results correlate better with
human judgment, giving a better Pearson correlation (0.357) as compared to BLEU (0.200), thereby showing ~70%
improvement over BLEU. Our research shows that, in the field of translation evaluation, existing resources like
back-translation and RDF could be useful.

Abstract

How can we effectively estimate the quality of translated texts in the medical field, where back-translation is
usually available and/or recommended for sensitive documents. This paper proposes a novel metric, GATE 1 ,
for translation quality estimation task, leveraging the Resource Description Framework (RDF) to encode both
semantic and syntactical information of the original and back-translated sentences into RDF graphs. The distance
between these graphs is measured to get the semantic similarity score to assess the quality of the translation.
Unlike traditional metrics like BLEU and METEOR, our approach is reference-less, capturing both semantic and
syntactical information for a comprehensive assessment of translation quality. Our results correlate better with
human judgment, giving a better Pearson correlation (0.357) as compared to BLEU (0.200), thereby showing ~70%
improvement over BLEU. Our research shows that, in the field of translation evaluation, existing resources like
back-translation and RDF could be useful.

Abstract

Named Entity Recognition (NER) is a use- ful component in Natural Language Process- ing (NLP) applications. It is used in various tasks such as Machine Translation, Summa- rization, Information Retrieval, and Question- Answering systems. The research on NER is centered around English and some other ma- jor languages, whereas limited attention has been given to Indian languages. We analyze the challenges and propose techniques that can be tailored for Multilingual Named Entity Recog- nition for Indian Languages. We present a hu- man annotated named entity corpora of ∼40K sentences for 4 Indian languages from two of the major Indian language families. Addition- ally,we present a multilingual model fine-tuned on our dataset, which achieves an F1 score of ∼0.80 on our dataset on average. We achieve comparable performance on completely unseen benchmark datasets for Indian languages which affirms the usability of our model.

Abstract

In the natural course of spoken language, individuals often engage in thinking and self-correction during speech production. These instances of interruption or correction are commonly referred to as disfluencies. When preparing data for subsequent downstream NLP tasks, these linguistic elements can be systematically removed, or handled as required, to enhance data quality. In this study, we present a comprehensive research on disfluencies in Indian languages. Our approach involves not only annotating real-world conversation transcripts but also conducting a detailed analysis of linguistic nuances inherent to Indian languages that are necessary to consider during annotation. Additionally, we introduce a robust algorithm for the synthetic generation of disfluent data. This algorithm aims to facilitate more effective model training for the identification of disfluencies in real-world conversations, thereby contributing to the advancement of disfluency research in Indian languages