detoxipy's People
detoxipy's Issues
[Utility Module] "Manual Features" Dictionary
Build a set of dictionaries containing "manual features" (domain context identified) capturing the following (Among others):
- Emojis
- Proper Nouns
- Punctuation
Build interfaces for extracting similar words using cosine similarity form embeddings vector matrix
Baseline Model @hyeness
- Naive Bayes linear model with TFIDF
Build baseline LSTM with processed text
Implement the test-train split using sklearn method
Function where this needs to be implemented: In build_dataset function inside nn_model_pipeline,
[Feature Creation Module] Embedding Feature Extractor
Integrate bias scoring into CV - [On Hold]
EDA Correlation of Text Features with label, identity, and comment type
Integrate unknown values into embedding code
EDA: Text
- Average word length
- Max word length
- Average char length
- Unique words vs. len
- Punctuations
- Emojis
Last-k words toxicity comparison
Pipeline architecture diagram
Building the evaluation framework architecture
Modules and pipeline scaffolds solution architecture:
- Data Preparation
- Feature creation - vectorising, ID predicting
- Model training
- Model Eval
[Feature Creation Module] BoW creator
Replicate other models on Kaggle @evsv
Integrate FT and W2V embedders with current NN code
EDA @ebago
EDA for Potential Data Preprocessing
- Punctuation vs. label
- Stop words vs. label
- Caps vs. label
- Proper nouns
Include identity word substitution in text processing & assess performance improvement
Integrate ID stripping into text prep module
[Dataprep/Feature creation Module] Spellcheck - Number of spelling errors
EDA: Non-Text
- Class Label Distribution
- Likes ratio: Disagreements vs. likes
- Toxicity concentration at end of comment
Train word embeddings for use in subsequent pipeline
Read about modelling approaches
Done- uploaded summary doc to repo
[Feature Creation Module] TF-IDF Creator
[Feature Creation Module] User defined category counter
Functions which, given user defined "categories" (punctuations, "upper case", "lower case", etc) counts both the raw number of occurrences and proportion in the dataset
DataPrep Module
- Tokenization
- Lower
- Remove punctuations
- Map punctuations/contractions
- Remove Stopwords
- Remove Whitespace
- Lemmatisation
- Identity words
- Proper Nouns
Preprocess data to get rid of identity words
Possible plan of attack:
- Heuristic based approach
- [?Explore] Embeddings based approach
- [Stretch] Adversarial Approach: https://github.com/yanaiela/demog-text-removal
Look into how we can repurpose previous policy labs code for non-NN code
[Feature Creation Module] Identity predictor sub-models - architecture design
A function to predict the presence of identity groups in a dataset
Implement early stopping into NN training
Find padding length as a function of maximum comment
Target function - in nn_model_pipeline -> build_dataset
Test and refine ID replacement regex
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. ๐๐๐
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google โค๏ธ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.