Machine Translation Digest for Sep 03 2025
Here is today's selection of cs.CL papers. The common theme across these papers is the enhancement of language processing models, with a focus on multilingual capabilities and pronunciation evaluation. Innovations include fine-tuning techniques for language models and the expansion of benchmarks to include lesser-known languages.
English Pronunciation Evaluation without Complex Joint Training: LoRA Fine-tuned Speech Multimodal LLM
This study demonstrates that a Multimodal Large Language Model (MLLM) adapted via Low-Rank Adaptation (LoRA) can perform both Automatic Pronunciation Assessment (APA) and Mispronunciation Detection and Diagnosis (MDD) simultaneously. Leveraging Microsoft's Phi-4-multimodal-instruct, our fine-tuning method eliminates the need for complex architectural changes or separate training procedures conventionally required for these distinct tasks. Fine-tuned on the Speechocean762 dataset, the pronunciation evaluation scores predicted by the model exhibited a strong Pearson Correlation Coefficient (PCC > 0.7) with human-assigned scores, while achieving low Word Error Rate (WER) and Phoneme Error Rate (PER) (both < 0.15). Notably, fine-tuning only the LoRA layers was sufficient to achieve performance levels comparable to those achieved by fine-tuning all audio layers. This research highlights that an integrated pronunciation assessment system can be established by adapting large multimodal models without full fine-tuning, utilizing a significantly simpler training methodology compared to previous joint models designed for simultaneous APA and MDD. This efficient LoRA-based approach paves the way for more accessible, integrated, and effective Computer-Assisted Pronunciation Training (CAPT) technologies for English L2 learners.
Expanding the WMT24++ Benchmark with Rumantsch Grischun, Sursilvan, Sutsilvan, Surmiran, Puter, and Vallader
The Romansh language, spoken in Switzerland, has limited resources for machine translation evaluation. In this paper, we present a benchmark for six varieties of Romansh: Rumantsch Grischun, a supra-regional variety, and five regional varieties: Sursilvan, Sutsilvan, Surmiran, Puter, and Vallader. Our reference translations were created by human translators based on the WMT24++ benchmark, which ensures parallelism with more than 55 other languages. An automatic evaluation of existing MT systems and LLMs shows that translation out of Romansh into German is handled relatively well for all the varieties, but translation into Romansh is still challenging.
LatPhon: Lightweight Multilingual G2P for Romance Languages and English
Grapheme-to-phoneme (G2P) conversion is a key front-end for text-to-speech (TTS), automatic speech recognition (ASR), speech-to-speech translation (S2ST) and alignment systems, especially across multiple Latin-script languages.We present LatPhon, a 7.5 M - parameter Transformer jointly trained on six such languages--English, Spanish, French, Italian, Portuguese, and Romanian. On the public ipa-dict corpus, it attains a mean phoneme error rate (PER) of 3.5%, outperforming the byte-level ByT5 baseline (5.4%) and approaching language-specific WFSTs (3.2%) while occupying 30 MB of memory, which makes on-device deployment feasible when needed. These results indicate that compact multilingual G2P can serve as a universal front-end for Latin-language speech pipelines.
Training LLMs to be Better Text Embedders through Bidirectional Reconstruction
Large language models (LLMs) have increasingly been explored as powerful text embedders. Existing LLM-based text embedding approaches often leverage the embedding of the final token, typically a reserved special token such as [EOS]. However, these tokens have not been intentionally trained to capture the semantics of the whole context, limiting their capacity as text embeddings, especially for retrieval and re-ranking tasks. We propose to add a new training stage before contrastive learning to enrich the semantics of the final token embedding. This stage employs bidirectional generative reconstruction tasks, namely EBQ2D (Embedding-Based Query-to-Document) and EBD2Q (Embedding-Based Document-to-Query), which interleave to anchor the [EOS] embedding and reconstruct either side of Query-Document pairs. Experimental results demonstrate that our additional training stage significantly improves LLM performance on the Massive Text Embedding Benchmark (MTEB), achieving new state-of-the-art results across different LLM base models and scales.
A Long Short-Term Memory (LSTM) Model for Business Sentiment Analysis Based on Recurrent Neural Network
Business sentiment analysis (BSA) is one of the significant and popular topics of natural language processing. It is one kind of sentiment analysis techniques for business purposes. Different categories of sentiment analysis techniques like lexicon-based techniques and different types of machine learning algorithms are applied for sentiment analysis on different languages like English, Hindi, Spanish, etc. In this paper, long short-term memory (LSTM) is applied for business sentiment analysis, where a recurrent neural network is used. An LSTM model is used in a modified approach to prevent the vanishing gradient problem rather than applying the conventional recurrent neural network (RNN). To apply the modified RNN model, product review dataset is used. In this experiment, 70\% of the data is trained for the LSTM and the rest 30\% of the data is used for testing. The result of this modified RNN model is compared with other conventional RNN models, and a comparison is made among the results. It is noted that the proposed model performs better than the other conventional RNN models. Here, the proposed model, i.e., the modified RNN model approach has achieved around 91.33\% of accuracy. By applying this model, any business company or e-commerce business site can identify the feedback from their customers about different types of products that customers like or dislike. Based on the customer reviews, a business company or e-commerce platform can evaluate its marketing strategy.
| Curated by yukajii.com |