So far we have uploaded data, vectorized it, clustered the data and we are ready to look for insights within the data.
There are various features on Relevance AI's platform for this purpose. The most recent, most sophisticated, most expansive and the easiest to use is the Explorer app.
On this page, we briefly overview the main functionalities of the Explorer app and how to share it with your peers.
For a more in-depth user guide please visit Explorer App.
They are all accessible on the main page and from the left-hand bar under Apps. Clicking on Explorer will open a page similar to the image below.
On the Explorer app, you can
- search within your data via word-matching and vector search
- filter and sort data based on various and configurable parameters
- configure different metrics and aggregations on your dataset
- design different views
- and many more
There are four main sections on this page. They are marked with A, B, C and D in the following images.
Placed on top, this section is to easily
- name your App
- share your app
- access the data page and data settings
- save your changes
- create a new visualisation of the data
It is the second section and is to
- configure search parameters for word-matching or vector search
- set up general and advanced filtering
- sort based on your data and your designed metrics
Metrics overview for clusters and data. This section provides you with a configurable and detailed analysis of categories in your data. Presentation, fields to consider, and type of analysis are all manageable through a few clicks.
For instance, in the image above, the dataset contains reviews and notes on a software called Spark. The data consists of fields such as user reviews and comments, number of time someone downloaded the software, number of thumb-ups, number of issues raised with the users, etc.
we can see that the reviews are categories into 10 groups. The "Spark and Machine Learning" category has the highest average on the number of likes (or starts as it is called on the dataset), whereas the second category, "Big data" has the highest average on forks (i.e. uses of the repository). The "Recommender System" has the least average on issues raised by users. It must be note that this configuration is 100% customisable. There are various features that one can employ to represent the insights; the image above is just a glimpse of what can be done on the dashboard.
The final section is where you can see data entries within their clusters/categories. The fields shown in each card and the aggregations on the right-hand side are all configurable.
In the image above, we are looking at the "Spark and Machine Learning" category.
All the metrics, that were set up (i.e. average number of stars, forks and open issues) in section C, are shown for every single category, in addition to number of reviews in each one.
Here, we have selected to only include "description" and "repo-name" as they are the two fields that we care the most about.
We are interested to know during what time period the Machine Learning category has been most successful. The chart illustrates that there is a boost from 2016 to 2017 in the number of received start; and not successful afterwards! A great starting point when searching for reasons behind success and failures!
When the configuration is finalized, make sure to save the last version using the button on the top right of the page. Click on the share icon, create a sharable link, copy the link which giver Read- access to whoever uses it (i.e. only the dataset owner or the person who is granted the write access can modify the page). The following images illustrate the process of generating a sharable link.
Updated 14 days ago