Major events:.

Polis and TTTC trial and description

Attempt to run TTTC on Google Colab

Prepare data and address related issues (encoding, NaN values, etc.)

Discussion of how to display and publish results

Discussion on how to manage FAQs

Reusable Know-how:.

Polis use:.

Onboarding is easy because you can try it out as an administrator simply by registering an account.

It is effective to experience the voting side in the sample and then try the admin side.

Using TTTC:.

Google Colab can be used to reduce the time and effort required to build the environment.

It is good to separate the analysis on your own data from the run on sample data.

Google Drive mounts are useful for persistence of results.

TTTC Data Preparation:.

When outputting CSV from Excel, care must be taken with encoding and data type.

Try to save in UTF-8.

Determine in advance how to process NaN values.

Error handling:.

Encoding-related errors can be checked and corrected with the appropriate editor (e.g., Cursor).

Errors due to LLM return format mismatches may be avoided by changing the model (e.g., from GPT-3.5 to GPT-4).

Result publication:.

Easily publishable using GitHub Pages.

Document Management:.

Choose a management location for information such as FAQs, taking accessibility and ease of updating into consideration.

A method of regularly transcribing Slack discussions into a structured document such as Notion is effective.

Tutorial design:.

It is easier to understand if you explain basic operations on sample data and analysis on your own data separately.

These know-how may be useful for similar projects and technology implementation in the future. In particular, useful insights have been gained on how to prepare and explain the work to a team that includes non-engineers.

next time (occasion)

--- chatting and stuff.

Save Point" concept to "save" projects that did not work out

---