Something I’ve been wrestling a lot with recently has been the amount of time I’ve focused on work and not on the people wanting to talk to me. There’re large stretches of time where I tend to be really busy and can’t reach out to the people I want to or even give them a response that I feel like is meaningful.

I wanted there to be a way where I could essentially buy myself a little bit of time where I could send interim messages that would allow me to answer and really respond when I had more to say.

So.

I had an idea. I wanted to make use of the data I had on myself that I managed to collect last week. My messages.

I had a rich amount of iMessages over 10 ish years that totaled to about 370 MB worth of text. All my messages back and forth responding to other people with my signature style of texting.

Make a chatbot using my own text messages.

What better training data for a conversational bot than your own text messages? I began by using the same data that I had from last week, and categorizing them by “text” and “Response”. After cleaning up the data a little I then attempted to find a machine learning model that could take in my formatted texts and use that as the training data.

After some digging I found this article that attempted to do something similar using doc2vec

Following the instructions and cleaning up the data gave me a chatbot where I could type in a real text that I got and see what the The chatbot version of me would say

I had some interesting responses from the chatbot that kind made sense like

Why am I so sad

I guess I could kinda look at this as me asking what movie they were thinking of watching.

Then I got some more non-sensical answers I found amusing.