I decided to make this project because I felt like asian history and asian voices are never heard in America. Before COVID, no one really understood the asian American experience and schools never studied our history. Asians are often viewed as the model minority and that we are next in line to becoming “white” so people assume we do not get discriminated. With that societal image of Asian americans, and the fact that we do not have enough power voices to speak up about our experience, society continues to discriminate. Especially with COVID, hatred against asian americans heightened. There are multiple instances of attacks against asians (atlanta shooting) and even video of attacks against asian grandpas. It is because Asian Americans spoke up now that we are all beginning to understand the Asian American experience more For this project, I decided to use speech recognition to mimic the idea of speaking up. The data from speech recognition would be used as a prompt to generate a poem.
Please create a .env file and paste in the key I sent to you in the comments Please check my requirements.txt for package versions that I used. I am currently operating this program on Python 3.10 rather than 3.11 for Tensorflow
python3 manage.py runserver
- Django
- Spacy
- Tensorflow
- Python Decouple
- The Fuzz
- Textstat
- Pyenchant
- Glob
- Numpy
- NRClex
- Please check requirements.txt for additional information
This system challenged me as a computer scientist because it made me think about purpose and intention in building systems. For this project, I initially had a lot of trouble coming up with an idea. However, I reflected on what mattered to me which inspired this project. Initially, I was afraid that the scope for this project was too huge and that I would not be able to complete it in time. Especially since I did not know how to work with Artificial Intelligence code or create a recurrent neural network model, I thought I would not be able to create what I wanted to create. Though, I did not give up and continued trying to learn and understand the code. Now, I know way more about A.I, how to build these text generation models, and the different tools that I can use. In addition to working with the neural network, I challenged myself by creating a fullstack website. I worked with Django 2 years ago before, but I forgotten most of it. However, I was glad that I pushed myself to relearn Django since I mostly focused on the frontend rather than fullstack.
- Full-FACE Poetry Generation (Simon Colton, Jacob Goodwin, Tony Veale)
- Acrostic Poem Generation (Rajat Agarwal, Katharina Kann)
- NLP based Poetry Analysis and Generation (Marmik Pandya)
The Acrostic Poem generation had a huge inspiration on my initial idea for the program. I wanted to also create an acrostic poem where we would ask the users to describe Asian Americans in adjectives. Using those adjectives, we would create an acrostic poem. However, it was difficult to do since I was not familiar with machine learning. This paper also inspired me to potentially add web scraping to this project, but I hesistated since I feel like poems on racism on the web are focused on certain races and not specifically Asian Americans. Overall, this paper helped me refine my purpose and intentions with what I am coding. In the full face poetry generation, they talked about corpus text generation and word similarity. Since the initial poem generated by my neural network is not the best, I decided to clean up the poem by word matching words that were not english. I created a word bank with all the words from the input training text file and use word similarity analysis where if the words match 80%, the non-english word would get replaced. If there was no match from the word bank, I turned to word suggestions. The poetry analysis inspired my evaluators. In the paper, the author talked about diction and affect on the readers. This inspired me to evaluate my poem on readability, and positivity on my poems. For the readability, I evaluated my poem using multiple formulas. My goal for readability is to make it as readable as possible so anyone can understand. The more readable, the easier it is for people to learn about Asian American history!
- Bjorn Ludwig
- https://www.poetryfoundation.org/collections/101589/asian-american-voices-in-poetry
- https://www.tensorflow.org/text/tutorials/text_generation
- https://pypi.org/project/textstat/
- https://dev.to/shantanu_jana/save-textbox-value-to-file-using-javascript-2ljp
- https://developer.mozilla.org/en-US/docs/Web/API/SpeechRecognition
- https://www.assemblyai.com/blog/javascript-text-to-speech-easy-way/
- https://poets.org/poem/muse
- https://poets.org/poem/how-i-got-name
- https://poets.org/poem/ramallah
- https://poets.org/poem/eden
- https://poets.org/poem/something-whispered-shakuhachi
- https://poets.org/poem/saignee
- https://poets.org/poem/leaving-seoul-1953
- https://poets.org/poem/white-boy-time-machine-override
- https://poets.org/poem/rootless
- http://www.kundiman.org/announcements/2020/3/20/10-poems-of-protest-to-read-this-march-5m22a
- https://yalereview.org/article/hades
- https://poets.org/poem/quarantine-0
- https://www.poetryfoundation.org/poems/89535/chinese-silence-no-22
- https://poets.org/poem/tanka
- https://poets.org/poem/dream-shoji
- https://poets.org/poem/i-ask-my-mother-sing
- https://www.npr.org/2021/04/12/985374483/today-i-am-a-witness-to-change-a-crowdsourced-poem-against-anti-asian-hate
- https://www.poetryfoundation.org/poems/58602/i-once-was-a-child
- https://www.poetryoutloud.org/poem/feasting/
- https://www.poetryoutloud.org/poem/after-the-dinner-party/
- https://www.poetryoutloud.org/poem/from-blossoms/
- https://www.poetryoutloud.org/poem/the-poem-youve-been-waiting-for/
- https://www.poetryoutloud.org/poem/slant/
- https://www.poetryoutloud.org/poem/the-emerald-mosque-on-the-hill/
- https://www.poetryoutloud.org/poem/filter/