nlp_a1
所属分类:自然语言处理
开发工具:Jupyter Notebook
文件大小:0KB
下载次数:0
上传日期:2024-01-25 16:39:25
上 传 者:
sh-1993
说明: 使用NLP单词嵌入模型的搜索引擎:SkipGram和GloVe。
(Search Engine using NLP word embedding models: SkipGram and GloVe.)
文件列表:
Code/
app/
A1_Engine_Search.pdf
# NLP Assignment 1 (AIT - DSAI)
- [Student Information](https://github.com/myothiha/nlp_a1/blob/master/#student-information)
- [Files Structure](https://github.com/myothiha/nlp_a1/blob/master/#files-structure)
- [How to run](https://github.com/myothiha/nlp_a1/blob/master/#how-to-run)
- [Dataset](https://github.com/myothiha/nlp_a1/blob/master/#dataset)
- [Evaluation](https://github.com/myothiha/nlp_a1/blob/master/#evaluation)
## Student Information
- Name: Myo Thiha
- ID: st123783
## Files Structure
- In the code folder, The Jupytor notebook files (training) can be located.
- The 'app' folder include
-- `app.py` file for the web application
-- Dockerfile and docker-compose.yaml for containerization of the application.
-- `template` folder to hold the HTML pages.
-- `models` folder which contains four model exports and their metadata files.
## How to run
- Run the `docker compose up` in the app folder.
- Then, the application can be accessed on http://localhost:8000
- You will directly land on the "Search" page.
## Dataset
- I used `brown` dataset (category 'News') from `nltk`.
## Evaluation
| Model | Window Size | Training Loss | Training Time | Semantic Accuracy | Syntactic Accuracy | Similarity (Correlation Score) |
|-------------------|-------------|---------------|---------------|--------------------|-------------------|-------------------|
| Skipgram | 2 | 10.16 | 0 min 03 sec | 0.00% | 0.00% | 0.08 |
| Skipgram (NEG) | 2 | 2.61 | 0 min 04 sec | 0.00% | 0.00% | 0.22 |
| Glove | 2 | 44.37 | 0 min 42 sec | 0.00% | 0.00% | -0.02 |
| Glove (Gensim) | - | - | - | 45.89% | 50.61% | 0.54 |
近期下载者:
相关文件:
收藏者: