Use surveillance techniques for unconventional ends












There is a conspiracy theory that said our phone is always listening to us. Whenever I talk about something it will soon be shown in advertisements on Instagram, Facebook, etc.
In order to show how it happened, I decided to surveillant one of my friend. I used live voice transcribe to record what Lisa said in three days.

Part of the text records

Then, I use A statistical parser to extract all the nouns in the record. It is a program base on JAVA. I use its Chinese version which is based on the Chinese Treebank.
In this way, we extracted thousands of words.
By using word2vec we extend the lexicon after de-duplication.
Word2vec is a group of related models that are used to produce word embeddings.
I record the first 10 ads I saw on each of the platform she used every day. 
I record the name and class of the ads.
Compared it with the lexicon I have got and I found out quite a lot of familiar.

Part of the ads records

As each of the web said most of the recommendations or ads are base on personal portray of Lisa.
She is an art student. Of course she like art exhibition. Of course, she need to learn python. Of course, she wants to go to Stanford. Of course she need to worry about here skin.
And there are also quite a lot of ads or recommendation seems quite ridiculous. And find relative keywords in the lexicon we got.




© 2020 by that's fine