Project information:

What is it:

Allai.Space targets inclusive language through three main pillars: vocabulary, sentiment, and syntax. For vocabulary, it examines common biases in language towards race, gender, and ability. The platform will directly replace key terms in typed or pasted text and produce it in a read-only output pane. Regarding sentiment, the web application utilizes the Google Cloud Natural Language program to examine overall sentiment of the passage and color the output pane accordingly. Using emotionally aware language is important for being inclusive not only in the interest of being positive, but also to avoid accidental offense. The final facet, syntax, also utilizes Google Cloud Natural Language. It identifies "entities", specifically people, and tracks any adjectives dependent on them. Based on the sentiment score of the adjective and the entity sentiment score of the person being described, suggestions will be generated in a passive, people-first format (i.e. disabled man will suggest modification to man with disability). Suggestions were generated using the Python Natural Language Toolkit library's Wordnet function.

Technologies Used:

  • Google Cloud Natural Language Processing APIs
    • Sentiment Score Analysis of text
    • Syntax Analysis - part of speech, dependency relation between words, entities
  • Python
    • Used for entire backend including NLP API calls
    • Used for find and replace of incorrect language with more inclusive counterparts
  • JavaScript, HTML, CSS
    • Used to develop an organized, clean, and easy to understand frontend
  • Flask Framework
    • Used to easily pass data (user inputted text and corrected version) between back and front end
  • JSON
    • Used for data storage to allow for data to persist even after closing app
    • Used for find and replace of incorrect language with more inclusive counterparts