Topic Report
The Topic report utilizes NLP techniques and LDA models to analyze and produce results that identify 10 topics. The 10 phrases that define the topic are shown as a bar chart to the right of the topic.
To the left of the bar chart is a list of calls that have been categorized under the topic. The list includes links to Wordbench for deeper analysis.
Top Level Filters
Organization and Folder:
These filters mirror your setup in Wordbench.
Topic analysis is undertaken at a folder level
Defaults to your first listed Organization and Folder
Select another Organization and Folder to view results for that Folder
Word type: only two and three word phrases are used for Topic analysis
Process Date:
Topic analysis requires a lot a data to ensure a meaningful result.
Therefore, Topic classification is undertaken over a rolling seven-day period, updating every 24 hours.
You can review the Topic classification from previous days by selecting a different process date.
Duration:
This filter will change the calls listed in the table.
It does not change the results of the topic analysis.
Topic Analysis
Dissects the 10 major categories based on number of calls in topic.
Hovering over any one segment will reveal the number of calls related to the topic and the percentage of calls within the dataset making up the topic.
Phrase Bar Chart
The bar chart to the left shows the top 10 phrases that define the topic. Each phrase is given a weighting between 0 and 1 to indicate its relevance to the topic.
Select a topic to show the phrases most relevant to that topic.
Table of Calls
As you cycle through and select a topic, the table will update to show the calls categorized under the topic.
The table consists of:
Filename
Duration (in minutes)
Topic number for each call
Probability that a call is categorized under this topic.
A probably less than 100% indicates that the call has been categorized under more than one topic. Calls with a probability of less than 50% are not shown.
Link to Wordbench for a deeper dive into that call
Copyright © VoiceAI Pty Ltd 2021, All Rights Reserved