08_Introduction_NLP_with_tensorflow - Disaster Example how the system knows based what we need to categorize the tweets.

[Irfanuzh]

In 08 Introduction to NLP with Tensorflow - Disaster example, how actually system knows we want to categorize tweets based on the disaster magnitude. Why not the system classify tweets based on something else like magnitude of love or hatred ...

If I am given these tweets and asked me to categorize manually , my first question will be 'based on what ?' .

[someshfengde]

you are providing labels associated with your data if you are labelling tweets such as love or hate ( binary classification ) then the model will be able to do that in disaster magnitude you have various types of levels associated with the tweets which helps model to determine the type of data model is getting in

[Irfanuzh]

The label, that is the ''target' column in traic.csv is telling 0 or 1 , not mentioning it is 0 when no disaster and 1 when disaster . it can 0 when *no love* and 1 when *love, *or 0 when *no hate* and 1 when *hate * etc ...

…

On Tue, Jul 27, 2021 at 12:54 PM Somesh ***@***.***> wrote: you are providing labels associated with your data if you are labelling tweets such as love or hate ( binary classification ) then the model will be able to do that in disaster magnitude you have various types of levels associated with the tweets which helps model to determine the type of data model is getting in — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#157 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AO34TGH3W4JX3TWETJYEC7DTZZNJ7ANCNFSM5BBRFWOQ> .

-- Regards Irfan

[mrdbourke]

If you'd like to create a new column based on different categories, you're more than welcome. What you put inside the column will depend on the information you'd like to add.

[Irfanuzh]

So what I understood is that given a collection of tweets and labels as follows to the model

text : Our deeds are the reason of this #earthquake .. , target :1
...
text : Just got send this photo from Ruby #Alaska.. .. , target :0

The model will learn by itself using the label values (target column) the binary classification is based on Disaster or Not.

So in this case there are two things the models will do

1. Identify based on what the classification is to be done
2. Classify the tweets based on the learned pattern.

Excuse me if I am asking dump questions .. I may have to go back to basics ..... :)

One last question, I asked all the above to ask this question .

Why don't we give a clue like we need to classify based on Disaster or Not as an input parameter to the model, as that will increase the accuracy and reduce the time of execution considerably .

[mrdbourke]

You're right!

Great breakdown.

As for giving a hint, we use the labels 0 & 1 as numerical forms of not disaster or disaster, a machine learning algorithm cannot handle text (it has to convert it to numbers first).