Sentiment Analysis Report

First, the speech of each speaker is checked against a predefined list of positive and negative keywords and normalised to calculate per-hundred count for both attributes. The analysed call is then compared with previous one hundred calls by default. However, the number of historical calls, to be compared with, is configurable to allow the comparison to be performed with higher/lower number of calls depending on the number of calls received in a day or a week and also on business type.

Please note, the number of previous calls to be compared with is by default set to 100. However, it can be configured to any number to represent the number of calls in one day or a week on an average received by your business.

It can also be configured to compare with previous calls that belong to the same category. For this, the call categorizer must be activated.

By default, the sentiment analysis module does not consider the category of the call while doing the comparison. If call categoriser is activated, each call can be compared with previous calls that belong to the same category. The functionality is desirable since the polarity of a call in a sales related conversation is expected to be different than that in a service call and the comparison should be performed among the calls that are of the same type.

For the sentiment score, each call is ranked first among all the calls in the comparison group based on the count of positive and negative words present in the speeches from each speaker. A voting algorithm is used to calculate the final ranking of the current call compared to other calls based on its rankings in both positive and negative list. The ranking is then converted into a sentiment score ranging from 0 to 10, where 10 represents maximum negative sentiment, 0 represents maximum positive sentiment, and 5 can be considered as neutral.

This approach allows you to identify more negative calls, even in a scenario where your business receives many negative calls during the time when the service is affected by some unexpected reason, for example, the server is down or any other issues and all the received calls would be considered containing negative sentiment in a normal situation.

Follow these instructions to access and analyse the Sentiment Analysis data for the selected service(s).

Step 1

To access the Sentiment Analysis report, log in to the Delacon portal and navigate to Reporting > Speech Analytics in the left sidebar of Delacon reporting portal and then select Sentiment Analysis. Alterntively, you can click on this direct link: https://pla.delaconcorp.com/siteui/reports/sentimentreport 

Step 2

To select the service(s) for which you want to access the report, click on Select Services. This will open a drop-down table where you can search and select services.

Fig 1: Select service(s)

Step 3

Select the data range using From and To data picker options.

Step 4

Select the Profile for which you want to generate the report

Fig 2: Select profile (if you have multiple sentiment analysis profiles)

Step 5

Click Submit to generate the report with selected Daily, Week, or Month.

Step 6

To change the data resolution, click on the intended button. The trend plot will be re-drawn to the desired resolution.

Please note, the selected profile for which the data is displayed in printed in the Select Profile section for your convenience.

The top part of the report provides trend of positive, negative, and default general calls for the selected duration and service(s) for both the speakers. Number of calls for all sentiment categories corresponding to the selected resolution (i.e. day/week/month) can be accessed by hovering the mouse pointer over the specific unit.

Fig 3: Example daily trend of calls with sentiment analysed

In the bottom section of the report, all the sentiment related data for both speakers are presented in tabular format. The table provides the functionality to filter the data based on different criteria as described below:

Fig 4: Drill-down data to short-list calls for further investigation

This table is essentially a tool for you to filter the calls out on which your business should concentrate to improve the service quality. You can filter and shortlist calls based on different criteria explained above.

Please treat the Call Id as number, whereas all other fields, including call duration, as decimal fraction for filtering purpose.

The data can be exported in CSV format for further analysis.

Please note, the number of previous calls to be compared with is by default set to 100. However, it can be configured to any number to represent the number of calls in one day or a week on an average received by your business.

It can also be configured to compare with previous calls that belong to the same category. For this, the call categorizer must be activated.