One-pager (Outline)

Required

• One-line, solution-agnostic description/title
  • SignIt! Is a language translator that takes in a text or file input and returns a language translated document in more readable terms. 

• Describe the opportunity in a solution agnostic way, such that more options could be brainstormed.

• • We need a way to make high level text easy to understand.

• Describe, briefly, the potential benefits in the customer’s words.
  • SignIt! will allow me to understand forms I receive from doctors, lawyers, and others easier, regardless of my education level and fluency in ny language.

• Summarize a mock customer case study involving this work (with quantifiable impacts).
  • A user who normally thoroughly looks at all his legal documentations were given our product to test how much faster his reading time improved and how thoroughly he was able to understand the document given translation product.
  • We conducted the mock experience using Wizard of Oz. The user was given a manually translated document with key phrases pertaining to numbers and binding agreements highlighted. The user was then evaluated based on his reading comprehension and reading time.
  • We concluded with the following result:
    • Users saved 50% time in reading
    • Users expressed significant improvement in ease of comprehension

• Top three areas of uncertainty, ranked by value of reducing uncertainty.

1. 1. Translation accuracy: product will simplified legal jargon convey the exact same meaning as the original text while supporting readability 
   2. Time Saving ability: how much time will the translation device actually save users
   3. Document privacy: ensure that model is not remembering private personal information during translation

• Current plan to explore those areas of uncertainty (link).
  • Progression towards resolving translation accuracy revolves around enhancing model’s translation and predictive power. We plan initially mannually identify key legal jargon, phrases, and other important numerical data in effort to speed up translation accuracy.
  • To explore the translator’s time saving ability, we plan to create side bar with highlighted phrases. This should allow users who do not want to read word-for-word translation save a large amount of time as they only need to glance over the key notes.
  • We have initilized steps for model to recognize words like “SSN”, “ID”, etc and not store the numerical numbers that directly follow that blank. This way our model can’t store your personal information but rather just copy pasting it over during a period of time.

• Summarize prior research related to this opportunity. What do we know?

1. 1. Quick and high quality text translator exists in DeepL (open API)
   2. Breaking down and paraphrasing complex text has been attempted, yet existing alternatives are not very effective
      1. Simplish.org attempts to paraphrase complex text by constraining the simplified version to a dictionary of 1000 common words
         1. The issue with this is that the text often became significantly longer while still not being very descriptive

• Leading signals we might observe if this is working (and not working).

• • Working:
    • The tool can translate text seamlessly between languages
    • The tool is able to successfully highlight important key terms in complex documentation
    • No important information is lost in the simplification process when paraphrasing
  • Failing:
    • Translation is slow and confusing
    • Highlighted key terms are irrelevant to what a user would want to know when signing a document
    • Text becomes significantly longer or more difficult to read after paraphrasing

• What other options did you consider (including benefits) (link)?

• • Complex B2B user interface:
    • Providing an extra interface for administrative users to oversee the different documentation users stored throughout the company
      • Companies using the tool are able to easily use the tool to oversee and monitor the many different transactions regarding legal documents that are happening  
  • Give users a Grammarly-like report on the content of the legal document
    • Provides an opportunity to educate users and give credence to the tool on why certain decisions were made in the simplification process
    • Difficult to implement, especially for the sake of MVP

• Why now, compared to alternatives? Really, truly, why now? Cost of delay (link)?

• • As students moving into the adult world, we have found that there are more and more legal documents we need to read, and be responsible for
  • In addition, there are more and more documents that are being signed online, which leads to less friction to delivering and signing documents
    • This could lead to mistakes being made while signing documents
  • The advent of covid has also lead to more remote work and behavior, which lends itself away from in person lawyers, and more online
• Brief summary of technologies involved (if known). Detail any areas of specialization required.

• • React
  • MongoDB
  • Some sort of framework to house our model, which will be written in python
    • Flask
    • NLP libraries (More testing needed)
  • NLP specialization is required

• Rough time frame (e.g. 1-3w, 1-3m, 1-3q, link).

• • 1-3hrs: Identify potential solutions to the problem.
  • 1-3days: Discuss the pros and cons of each solution and choose one.
  • 1-3weeks: Do user research to identify customer needs and what features need to be in the application.
  • 1-3months: Develop hi-fi prototype, text, and start the development of the web application to do basic language translation. Begin training model. Use a human in the workflow to catch errors. 
  • 1-3quarters: Continue training model and have customers reject simplifications.
  • 1-3years: Expand the app out for hierarchical business models.
  • 1-3decades: Lease the tool out for larger companies that perform translation on a regular basis.