Build a prototype question answering system

[Colin-Codes]

Question - Question matching using basic machine learning

• [x] Construct a dummy data set with questions and answers (many questions onto each answer)
• [x] Build a simple knn model to match new questions on existing answers via old questions

[Colin-Codes]

Must decide what the something more convoluted will be....

linguistic dependency approach? A neural network? Another kind of algorithm? Comparison of vector scores between Qs and As Data mining of docs

[Colin-Codes]

Build a GRU:

https://medium.com/mlrecipies/deep-learning-basics-gated-recurrent-unit-gru-1d8e9fae7280

[Colin-Codes]

Build an LSTM:

https://adventuresinmachinelearning.com/keras-lstm-tutorial/

[Colin-Codes]

OR:

https://towardsdatascience.com/nlp-building-a-question-answering-model-ed0529a68c54

(pull the repo from GitHub) Too complicated

[Colin-Codes]

• [x] Build a hello world email responder app (Description:https://repl.it/talk/learn/How-to-Make-a-Python-Email-Bot/8194/20686) (hal9000.origin@gmail.com)
• [x] use pyiinstaller to package it (did cause a firewall digital signature warning)
• [x] build a sample training set
• [x] Build an RNN to tailor the response based on intent (https://medium.com/mlrecipies/deep-learning-basics-gated-recurrent-unit-gru-1d8e9fae7280)
• [ ] May need to consider data augmentation (https://machinelearningmastery.com/how-to-configure-image-data-augmentation-when-training-deep-learning-neural-networks/)
• [ ] Consider handling uncertainty

Used this as a basis:

https://codekansas.github.io/language.html

[Colin-Codes]

Done