Thanks for the new datasets, I saw Quentin Tarantino curse words and deaths and had to check it out!
The default viz (after loading the data) uncovered an interesting pattern where Quentin is using a particular group of curse words in different movies -
I was curious what other vizs would uncover so I used the filter on the right side, but I didnt feel it moved as quickly as I would like. I’m not sure the fix, but arrow keys (up/down) might be nice to switch through choices in each category ( e.g. In SORT category - “First to last”-> “Last to first” → “Smallest to largest” → “Largest to smallest”) and then another down click would jump into the next category of filter choices ALIGN / min → center → etc.
Alternatively, a browse mode, which might put 4-6 in one screen and then let me hit the right-left arrow keys to flip through the group would greatly speed my ability to quickly surmise what vizs might be interesting and offer insights.
This last thought prob. doesnt make sense to more seasoned viziers (is that a word? I think alladin ): It might be insightful for some to see a list of all possible column combinations (for tables with low cardinality) and even those with higher cardinality by selecting particular columns. Sort of like a cartesian product of the visualizations available for a particular set of columns, then if you want to add the gravy, a way to flip through the vizs so I can see all the viz options available. So, this way I can visually walk through all variations of my X-Y column setup and then walk thru all the options available for charting. I dont imagine it would be hard to do for one small table, some work but not overly complex, but the benefit it would give would be great for ppl who otherwise would manually need to go through different options. If I could check multiple columns in a table, then checkmark a few different viz types, and then click “Generate” to create a page that would let me click-thru and see the results for a sample set of data for each one–I believe that would have some value to business users.
Hey @dannylee8,
I’m glad you’ve had some fun with Tarantino’s obscenity
Thanks for the tip about cycling through the configuration options on the right side of the chart with the arrow buttons.
What exactly would you like to put 4-6 of in one screen in the suggested “browse mode”? Charts?
I love the idea of showing a collection of potentially interesting charts using combinations of the data series. We have this idea for some time and will work on it later. This could speed up the exploration of a new dataset like a charm.
Yep 4-6 charts, but I was picturing an explorer type functionality. So if I have 4 columns as potential X axes and 5 columns as potential Y axes, then I can checkmark them in a list and then get 20 variations. If I also select 3 different types of graphing options, I would get 60 combinations.
Then I can flip through the pages of the result set in groups of 4-6 charts and look for any new ideas or viewpoints.
Yes, The Voyager and Voyager 2 pages are very interesting – its seems like a way to quickly visualize many more possibilities than can fit in my poor memory/imagination!
It kind of makes me think of a presentation I saw from zilliz (vector db company) that was talking about using LLMs to come up with better questions. So, we ask one LLM a question and get a response. Then we that the response and use it as a context input to other LLMs and as the LLM to come up with interesting questions that would generate this kind of data.
So, by inputting a set of data we want to explore and letting the system come up with the variations for us we are able to explore ideas about the data we may not have thought about – even if logically speaking, one might think of putting two particular columns of data and comparing them against one another.
However, I would not be surprised if this doesnt make as much sense to seasoned statisticians and those able to visualize the possibilities in their mind, but for the other population of users who don’t have as much knowledge or experience it can open up new ideas and possiblities.
I think such a feature would make a ton of sense, and we’ll implement it at some point in the future.
Based on my interviews with data scientists and analysts, such a feature would be handy for them when they start working with a new dataset, so I don’t think this would only be used by the less stat-savvy users. We’ll see
