Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/davidsvy/Neural-Scam-Artist

Web Scraping, Document Deduplication & GPT-2 Fine-tuning with a newly created scam dataset.
https://github.com/davidsvy/Neural-Scam-Artist

dataset deduplication fine-tuning fraud gpt2 huggingface lsh minhash nlp pytorch readability scam transformer web-scraping

Last synced: 3 months ago
JSON representation

Web Scraping, Document Deduplication & GPT-2 Fine-tuning with a newly created scam dataset.

Host: GitHub
URL: https://github.com/davidsvy/Neural-Scam-Artist
Owner: davidsvy
License: mit
Created: 2021-10-18T05:55:18.000Z (about 3 years ago)
Default Branch: master
Last Pushed: 2021-10-30T15:57:53.000Z (about 3 years ago)
Last Synced: 2024-05-15T04:34:02.087Z (6 months ago)
Topics: dataset, deduplication, fine-tuning, fraud, gpt2, huggingface, lsh, minhash, nlp, pytorch, readability, scam, transformer, web-scraping
Language: Python
Homepage:
Size: 188 KB
Stars: 23
Watchers: 2
Forks: 1
Open Issues: 0
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

        


  Neural Scam Artist






      

        

       

        

       

        



TL;DR\

A dataset of scam emails is scraped from an anti-fraud website. The dataset is then deduplicated

using MinHash and LSH. The deduplicated dataset is used for fine-tuning GPT-2.



  





  Comic stolen from Agent-X Comics.

  



:book: Table of Contents

===

  * [➤ Project Description](#project-description)

    * [➤ Objective](#objective)

    * [➤ Web Scraper](#web-scraper)

    * [➤ Deduplication](#deduplication)

    * [➤ GPT-2](#gpt-2)

  * [➤ Shared Files](#shared-files)

  * [➤ Requirements](#requirements)

  * [➤ Installation](#installation)

  * [➤ Usage](#usage)



:cloud: Project Description

===



Objective

---

The goal of this project is create a new dataset of fraudulent emails that can advance the

research on intelligent email assistants.



Web Scraper

---

Data is scraped from the website [https://antifraudintl.org/](https://antifraudintl.org/). 

At first, a set of thread urls is collected and stored. Then, each thread is searched for 

emails. For each thread, at most one email is kept as the rest are duplicates. Metadata 

(Subject, Date etc) is removed. The resultant dataset is stored inside a csv file.



Deduplication

---

To avoid the quadratic complexity, a cheap alternative is selected: MinHash and LSH using the [datasketch library](https://github.com/ekzhu/datasketch). For each document, this method 

efficiently locates its nearest neighbors. Because this leads to a a large amount of false

negatives (i.e. dulpicate documents that are classified as non-duplicates), the approach is

extended by creating a duplicate graph. Nodes in this graph represent documents and are connected

with an edge if their respective documents have been classified as duplicates. To deduplicate the 

dataset, [connected components](https://en.wikipedia.org/wiki/Component_(graph_theory)) of the 

graph are located and for each component only a single node is selected. A 

[readability criterion](https://en.wikipedia.org/wiki/Readability) is used for selection.



GPT-2

---

A small pretrained GPT-2 model from the 

[Huggingface library](https://huggingface.co/transformers/model_doc/gpt2.html#gpt2lmheadmodel)

is fine-tuned on the deduplicated dataset. A collection of ~~cherry-picked~~ randomly selected 

generated samples can be found here [here](https://github.com/davidsvy/Neural-Scam-Artist/blob/main/generated_samples/generated_samples.txt).



:file_folder: Shared Files

===

| Resource | Size | #Samples | Link |

|-------------------|---|---|---|

| **Full dataset**          | 128.5 MB  | 85,160  | [Link](https://drive.google.com/file/d/1CoZp1F0FqB3pOqYlQ7X9XqCChppZFUps/view?usp=sharing)  |

| **Deduplicated dataset**  | 74.2 MB   | 58,227  | [Link](https://drive.google.com/file/d/19JXTPTqV9gaKzHqGdbyyEuEfXD2l5DCc/view?usp=sharing)  |

| **Thread urls**           | 6.4 MB    | 95,324  | [Link](https://drive.google.com/file/d/1AmVIqCnWzSCqexTv02wOBAnPhiTkgHkP/view?usp=sharing)  |

| **GPT-2 Checkpoints**     | ~1.5 GB   |   | [Link](https://drive.google.com/drive/folders/1RUV2gPbGUetBFlIJZ9_W-ARB70x_9s-L?usp=sharing)  |



:toolbox: Requirements

===

See `requirements.txt`.



:gear: Installation

===

```

$ git clone https://github.com/davidsvy/Neural-Scam-Artist

$ cd Neural-Scam-Artist

$ pip install -r requirements.txt

```



:roll_of_paper: Usage

===

To generate dataset (~3 hours on Colab):

```

$ python create_dataset.py [-c configs/create_dataset.yaml]

```

To deduplicate dataset (~30 minutes on Colab):

```

$ python deduplicate_dataset.py [-c configs/deduplicate_dataset.yaml]

```

To train GPT-2 (~3 hours/epoch on Colab with K80):

```

$ python gpt2_train.py [-c configs/gpt2_train.yaml]

```

To generate text with GPT-2:

```

$ python gpt2_sample.py [-c configs/gpt2_sample.yaml]

```