A poetry generator from a scrapped corpus of Spanish poetry. EDA and general NLP tasks are included.
Here a visualization of the generator performance with streamlit. Given some words.... it generates a beautiful poem in Spanish✨✨✨✨
We generated an overview of the whole data. We analyze the scope and length of the vocabulary involved, generating some nice visualizations ☁️☁️☁️
We decided to make some word counts as well as search for relations between authors and poems in the whole dataset 📈
We also took into account specific authors and established some comparisons. We detected relations between textual data such as antithesis and polysemy. Awesome isn't it? 🤩
An embedding model was build to detect polysemy, similar words, and common word collocations in poetry. So many word relations in poems!
.
Also, Voronoi graphs were made...📈📈📈📈
.
-
Script for scrapping the data: script that generates a .csv files with the poems scraped from the blog entries of a Spanish poetry webpage (link to the webpage here).
-
Poem generator visualization (Streamlit): poems generator code using streamlit.
-
[EDA of the poetry dataset](notebooks/data exploration.ipynb): Exploratory Data Analysis of the dataset, including a basic NLP complete task!
-
[Poem genetator code](notebooks/poetry generator.ipynb): code to generate synthetic poems with a RNN.
- Recommended: open the notebooks in Colaboratory.
This project has been presented as a talk in the PyConEs 2020 (Pandemic Edition). You can find the slides in this repo and the video in youtube.