This research introduces KoGEC, a Korean Grammatical Error Correction system using pre\--trained translation models. We fine-tuned NLLB (No Language Left Behind) models for Korean GEC, comparing their performance against large language models like GPT-4 and HCX-3. The study used two social media conversation datasets for training and testing. The NLLB models were fine-tuned using special language tokens to distinguish between original and corrected Korean sentences. Evaluation was done using BLEU scores and an "LLM as judge" method to classify error types. Results showed that the fine-tuned NLLB (KoGEC) models outperformed GPT-4o and HCX-3 in Korean GEC tasks. KoGEC demonstrated a more balanced error correction profile across various error types, whereas the larger LLMs tended to focus less on punctuation errors. We also developed a Chrome extension to make the KoGEC system accessible to users. Finally, we explored token vocabulary expansion to further improve the model but found it to decrease model performance. This research contributes to the field of NLP by providing an efficient, specialized Korean GEC system and a new evaluation method. It also highlights the potential of compact, task-specific models to compete with larger, general-purpose language models in specialized NLP tasks.

General comments: The paper is well written. I would just remove "cryptographic" in the abstract. Not a lot of people are familiar with differential privacy and many of them would get confused with the notion of cryptography in which there exists an algorithm that reverts the "encrypted" data to its original form (I consider that once the data is noisified, its original form is essentially "lost"). Technical comments: Following the argument in the supplementary information of ref. [30], the authors argue that being able to find a perfect SSC clustering breaches the notion of differential privacy (DP). Consider the following problem: Name: SSC approximation.

What do we do with archival images in the time of AI image creation? As I have been writing this article, Dall-E and similar projects have become extremely popular. I had a chance to access Midjourney, a platform similar to Dall-E.

Capturing snowflakes on film can be quite the feat, as photographers have mere before the tiny ice crystal's intricate details melt – but a new video shows the event in reverse. Photographer Jens recently shared a stunning video showing already melted snowflakes freezing back to their original form. Each shot begins with a small droplet of water that begins to sprout icicles until it returns to the unique design. The movie was done using highly detailed macro photography, which is capable of making very small object look larger than life size. Capturing snowflakes on film can be quite the feat, as photographers have mere before the tiny ice crystal's intricate details melt – but a new video shows the event in reverse.

Stanford NER is a named-entity recognizer based on linear chain Conditional Random Field (CRF) sequence models. This post details some of the experiments I've done with it, using a corpus to train a Named-Entity Recognizer: the features I've explored (some undocumented), how to setup a web service exposing the trained model and how to call it from a python script. Once Java is setup, you can run Stanford NER using one of the already trained models, which are distributed together with the zip file. Create a file with a sample sentence in english. Then, to apply the english.all.3class.distsim.crf.ser.gz to the sentence above, run the java command shown bellow: This section describes the basic steps to train your own NER model, from pre-processing the corpus (if needed), creating k-folds for cross-fold validation, defining the features to use, and running Stanford NER in evaluation mode.