## Tech

## Conversations Are Not Flat: Modeling the Dynamic Information Flow across Dialogue Utterances

Let me introduce an interesting paper I have recently read regarding dialogue modeling and evaluation.

## DialogueSentenceBERT: SentenceBERT for More Representative Utterance Embedding via Pre-training on Dialogue Corpus (2)

Previous post: https://songstudio.info/tech/tech-38

## DialogueSentenceBERT: SentenceBERT for More Representative Utterance Embedding via Pre-training on Dialogue Corpus (1)

In the 2nd quarter of this year, I conducted research on developing a contextualized sentence embedding model, which is more optimized on an utterance embedding for dialogue tasks.

## Averaging methods for F1 score calculation in multi-label classification

There are various evaluation metrics to test the model we trained when conducting machine learning projects.

## DialoGLUE: A Natural Language Understanding Benchmark for Task-Oriented Dialogue

It’s been a long time since I’ve posted a review on an NLP paper.

## Multi-turn chatbot project (3): GPT-2 chatbot with multi-turn generation settings

(This post was modified in December 2nd after the re-implementation of the project to prevent you from confused by the difference between the repository and the contents of the post.)

## Multi-turn chatbot project (2): Transformer chatbot with the ReCoSa structure

Following the introductions that I posted last time, today let’s talk about the transformer model using the **ReCoSa**(the Relevant Contexts with Self-attention) structure, which is the first model for the multi-turn chatbot.

## Multi-turn chatbot project (1): Introductions

Starting with this, I’m gonna post about the personal project developing the **“Multi-turn chatbot”**.

## Computing Machinery and Intelligence

Few of those majoring in computer science would not know **“Alan Turing”**.

## Bias-Variance trade-off

“Bias-Variance trade-off” is one of the fundamental concepts in Machine Learning studies, which means that there is a trade-off relation between two errors(or losses), bias and variance when evaluating the generalization capacity of ML algorithms.

## Revisiting the transformer NMT project

It’s been a while.

## The Curious Case of Neural Text Degeneration

This post is the review of the publication, *Holtzman, A., Buys, J., Du, L., Forbes, M., & Choi, Y. (2019). The curious case of neural text degeneration*.

## Beam Search

**Beam Search** is a tree search algorithm based on “Best First Search” method used in various NLP tasks frequently.

## Monty Hall problem

“Monty Hall problem” is a very famous mathematical problem related to conditional probabilities.

## BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

**BERT** stands for “Bidirectional Encoder Representations from Transformers” which is one of the most notable NLP models these days.

## Neural Machine Translation with Transformer in Pytorch

We reviewed the famous *Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., … & Polosukhin, I. (2017). Attention is all you need. In Advances in neural information processing systems (pp. 5998-6008)* last time.

## Attention Is All You Need

This post is about the famous **Transformer**, which has advanced the progress of NLP research.

## Union-Find Algorithm

## Attention mechanism in simple RNN based model

Since the basic idea seeing the overall input contexts as references is same, it is obvious that we can use attention to basic RNN based models.

## Attention mechanism in seq2seq model

**Attention** mechasnism is one of the most important concepts in NLP field.

## Sequence-to-Sequence model

**Sequence-to-Sequence(seq2seq)** model is a Deep Learning model usually used for machine translation, text summarization etc.

## 2019-2 YBIGTA CONFERENCE

YBIGTA is the Big Data Academy in the college of Engineering, Yonsei University.

## AICON 2019

It has been quite a long time, but I finally managed to post about AICON 2019, which is a global AI conference held at Yangjae R&D Innovation Hub, in Dec 17th, 2019.

## Minimum Spanning Tree

**Spanning Tree** is a tree that has all vertices from an original graph and has a minimum number of edges.

## Mean Square Error & Cross Entrophy Loss

The most commonly used loss functions in Machine Learning/Deep Learning are **Mean Square Error** and **Cross Entropy loss**.

## Traveling Salesman Problem

**Traveling Salesman Problem(TSP)** is one of the most famous problems in algorithm and the basic example is below.

## Finding the Diameter of a Tree

## Operator Overloading

Simply saying, this is about re-defining basic operators for primitive types and it can be used for operating between objects from classes or structs.

## Priority Queue

I found the data structure called “Priority Queue” in `<queue>`

libaray.

## Segment Tree

## Default values in a map

## Making Pytorch custom dataset

When doing simple practices, we can download datasets provided by the framework itself, process them into loaders and put them into our models.

## Thoughts about time and memory

## The difference between DFS and BFS

Today I solved a problem which should be approached with BFS.

## Runtime error

During these several days, I have suffered from runtime errors while solving Baekjoon Online Judge problems.

## Using partial sums

## Updating the value in map

I have usually used `map.insert()`

to put a value in the map and known that this would update automatically if there is already the existing key value.

## Finding cycles in graph application

Until now, I have thought that I have become accustomed to solving problems clearly with graph searching, for instance, DFS, BFS. But I realized that a problem that doesn’t seem to be solved with a graph can be processed with a graph structure.

## Binary search

Binary search is a powerful method to find the desired value in a sorted sequence, but I have not been used to decide whether the given case should be solved with Binary search.

## next_permutation

I noticed that there is a function to get permutations of a given sequence.

## DailyLife

## I am starting my master’s program in Fall 2022!

Among the several admitted options, I chose to go for the **Master of Science in Engineering in Computer and Information Science (MSE in CIS)** program at the **University of Pennsylvania**, PA in Fall 2022.

## I am currently a ML engineer at Mindlogic!

From October, 18th, I have been working as a Machine Learning engineer in Natural Language Processing at **Mindlogic, Inc.**, Seoul.

## Starting an internship at Seoul National University!

I’m happy to announce that I’m currently participating in the internship at **Machine Intelligence lab(MILab), Seoul National University** from June 1st.

## Lessons from Boostcamp AI Tech

It’s been a while.

## Am I a good researcher?

I feel so regretful…

## The time has come for me to worry about the body.

I’ve worked out too much to lose weight recently.

## I have to prepare for two exams.

I have decided to prepare for TOEFL these days since my previous score was expired last summer.

## Fresh new start!

It’s been a while!

## It’s been a while…

This is almost a month of posting.

## My laptop is almost dead

My laptop which I bought right after the termination of military service and has used for about 1 and a half years is dead.

## Transferred to the custom domain!

Finally, Songstudio is successfully transferred to the custom domain.

## Intership final presentation/interview over

Yesterday, the final presentation and official interview of LG CNS internship finally finished.

## Assgined as the research intern in LG CNS AI/BIGDATA Techolony Team Language part

I already heard that I was assigned to the AI research team, but I was worried because there are many specific research parts in the team so I had no idea about which part I’m going to.

## Working out again

Today, I started to exercise again by registering the school’s fitness center.

## To do list during the summer vacation

These are works to do during this vacation.

## Blog starts!

SongStudio is now open today!