If we upload the mock data set shown at the bottom, to The Explorer then you will receive a Visual Dashboard - note it is a mock set because it is small and it simplifies the calculations that are done to generate the graph data that is visualized in the Dashboard. Here is the URL to the Dashboard to see it yourself: https://explorer.gavagai.io/dashboards/305390/public?token=vmAblZwWAe
Here we can see a X-Y Scatter plot of Topics found from The Explorer. The X axis is the score in Percentage Terms how often a Topic was mentioned from the text data. The Y axis is the "Grade" score from all the reviews mentioning a specific Topic and finding the average grade for those reviews mentioning that Topic. We circled the Topic Rooms which has a Topic Occurrence of 60% (12/20) and a Topic Grade of 4.5. If you scroll to the bottom to manully inspect the data set we can see that indeed there were 12 reviews mentioning the Room Topic (The Explorer would be the one doing this for you). With those 12 reviews we can add up their grade:
4
5
5
5
5
5
4
5
3
4
5
4
Which is 54. Then divide it by 12 which is 4.5. Thus we found the score related to the Topic Room's average Grade. This is repeated for all the Topics the Explorer finds and is plotted on this graph. Note The Explorer attempts to find all the Topics in each review and gives us a list of the most prevalent 30 Topics on default. The dotted lines represent the average scores for the Topics. You can also click the bubble to be taken to the Topic Page where you can see the raw scores and text examples.
By clicking any Topic you are brought to a new View focused on that Topic. This graph checks the file you upload and if there are Dates in a column which holds Dates or Timestamps for each review. And if there are Dates then this graph will appear and show bars and lines on a graph. The white bars represent the Topic's Occurrence per day/month/year depending on the type of dates there is in the data. The blue line represents the Topic's sentiment score. Please check here to learn more about our Sentiment Scoring system. https://docs.gavagai.io/topics/#sentiment-analysis
(note this graph is from another project because our original data did not have a Date column)
By clicking the Topic Room you are brought to a new Topic focused view on the Topic Room. If you scroll down slightly you will see a graph labelled Topics Related to Rooms. This is a bar chart that breaks down other Topics related to the Topic Room in the reviews mentioning the Topic Room. That is, when the Topic Room is mentioned what other Topics are mentioned in those reviews. Here we can see that of the Room mentions, about 40% of the reviews are mentioning the Topic Staff.
From the X-Y Scatter plot graph, we can see some bubbles that are higher on the y-axis that is, their Topic's Grade average is higher than other Topics. Thus, we call those Important Topics. These bars represent how much they deviate from average grade.
Related Topics to the Important Topics shows a graph where each bar represents a related Topic to the original Topic. In this case the Topic Breakfast has the related Topics Wifi, Stuff, Happy, Hour. The height of the bar represents the difference from the Topic Breakfasts average grade in percentage. That is, The Explorer finds all the reviews mentioning both Topic Breakfast and the Topic Wi-Fi and calculated the average grade for the reviews mentioning both Topics. Then finds the percentage difference from the Topic's Breakfast average grade.
Here we can see a bar chart where the white bars represent the Topics the Gavagai Explorer has found and the height or Y axis represents the Percentage of reviews found mentioning that Topic.
By toggling the big white pill button, and turning it blue, we now see blue bars that represent the Topic's average grade. Note the y-axis is cut and might not start from 0.
The above graphs were possible because for each review, the person who wrote the review also left a grade score or "quantitative score" that is a number on a scale to represent their satisfaction of the hotel. If there was no "quantitative score" then the Explorer would default to using Sentiment found from the text data.
Search words:
explain the dashboard, what do the graphs mean, the graphs