Working with reports
You can use SpeechSense reports to analyze dialogs between agents and customers and evaluate agents' performance. Build a report:
- Evaluation form: To analyze multiple evaluation parameters.
- Comparison: To compare multiple evaluation parameters over time.
- Details: To analyze changes in a single evaluation parameter over time.
If you want to use a dialog feature, e.g., reason, topic, or dialog outcome, as a filter or evaluation parameter, use a semantic attribute in your report.
Building an Evaluation form
-
Open the SpeechSense home page
. -
Go to the space you need and select a project.
-
Go to the Reports tab.
-
If the project has no saved reports yet, select Evaluation form. If it already contains reports, click Create report → Evaluation form.
-
In the window that opens, click the Period field and set the start and end dates. The report will only include dialogs from within the specified period.
-
Optionally, if you want the report to include dialogs with specific properties, do the following:
-
Click
next to the Period field. -
In the drop-down list, select a dialog property, e.g.,
AgentorProduct. -
Depending on the selected property, enter its values, comma-separated, or specify a range of values.
Tip
To filter dialogs by semantic properties, such as topics or totals, use a semantic attribute as the filter.
The report will only include dialogs which meet the filter conditions. To use multiple filters, apply them one after another. For more information, see Filtering in the report.
-
-
Click Group by. In the drop-down list, select the metadata field to group your report data by. For more information, see Filtering in the report.
-
Click
Add parameter. -
Set up the evaluation parameter:
-
In the Evaluation parameters field, select the appropriate parameter and specify its value or a range of values.
Evaluation parameters are used as criteria for assessment of agent performance. For more information about the parameters and how to generate a report based on them, see Evaluation parameters.
If you want to use a dialog feature, e.g., reason, topic, or dialog outcome, as an evaluation parameter, use a semantic attribute.
-
Under Weight, specify a percentage.
Weight indicates how critical a parameter is for agent performance evaluation. The higher the parameter's weight, the greater its value in the report. The total weight of all evaluation parameters must equal 100%. For more information about weight and its use, see Evaluation parameter weight.
-
(Optional) In the Parameter name in report field, specify the name that will be used to caption the parameter on the chart and in the table. By default, this name matches the Evaluation parameters field value.
-
-
If you need to add more evaluation parameters, click
Add parameter. -
Click Create report.
-
To reuse the report, click Save report. Specify its name and, optionally, a description.
The report will be available in the selected project in the Reports tab.
Switching between an Evaluation form and dialogs
A report enables you to directly switch to the list of dialogs that meet the criteria specified in the report. This way, you can get more details about dialogs and analyze their specifics.
To switch from a report to a dialog list:
- Create a report Evaluation form.
- In the report chart, click the rectangle of the evaluation parameter you need. This will open a list of dialogs according to the active filters and the parameter you selected.
- Go to the dialog of your choice. Its page will show detailed info about the dialog, e.g, the agent's performance data in the form of dialog summary, agent and customer tags, transcript of the audio recognized using SpeechSense.
Building a Comparison report
-
Open the SpeechSense home page
. -
Go to the space you need and select a project.
-
Go to the Reports tab.
-
If the project has no saved reports yet, select Comparison. If it already contains reports, click Create report → Comparison.
-
In the window that opens, click the Period field and set the start and end dates. The report will only include dialogs from within the specified period.
-
Optionally, if you want the report to include dialogs with specific properties, do the following:
-
Click
next to the Period field. -
In the drop-down list, select a dialog property, e.g.,
AgentorProduct. -
Depending on the selected property, enter its values, comma-separated, or specify a range of values.
Tip
To filter dialogs by semantic properties, such as topics or totals, use a semantic attribute as the filter.
The report will only include dialogs which meet the filter conditions. To use multiple filters, apply them one after another. For more information, see Filtering in the report.
-
-
Click Group by. In the drop-down list, select the metadata field to group your report data by. For more information, see Filtering in the report.
-
Click
Add parameter and select the one you need.For more information about the parameters and how to generate a report based on them, see Evaluation parameters.
-
Specify what data you want to get: total of evaluation parameter values, average, minimum, or maximum value of the evaluation parameter.
In the Comparison report, you can only select numerical evaluation parameters. If you select a tag as a parameter, the total, average, minimum, or maximum number of times the tag was assigned to dialogs over the specified period will be counted.
-
To add more evaluation parameters, click
Add parameter. -
Click Create report.
-
To reuse the report, click Save report. Specify its name and, optionally, a description.
The report will be available in the selected project in the Reports tab.
Building a Details report
-
Open the SpeechSense home page
. -
Go to the space you need and select a project.
-
Go to the Reports tab.
-
If the project has no saved reports yet, select Details. If it already contains reports, click Create report → Details.
-
In the window that opens, click the Period field and set the start and end dates. The report will only include dialogs from within the specified period.
-
Optionally, if you want the report to include dialogs with specific properties, do the following:
-
Click
next to the Period field. -
In the drop-down list, select a dialog property, e.g.,
AgentorProduct. -
Depending on the selected property, enter its values, comma-separated, or specify a range of values.
Tip
To filter dialogs by semantic properties, such as topics or totals, use a semantic attribute as the filter.
The report will only include dialogs which meet the filter conditions. To use multiple filters, apply them one after another. For more information, see Filtering in the report.
-
-
Click Group by. In the drop-down list, select the metadata field to group your report data by. For more information, see Filtering in the report.
-
Click
Add parameter. -
Select the evaluation parameter whose changes you want to view.
For more information about the parameter and how to generate a report based on it, see Evaluation parameter.
-
Specify what data you want to get: total of evaluation parameter values, average, minimum, or maximum value of the evaluation parameter.
In the Details report, you can only select numerical evaluation parameters. If you select a tag as a parameter, the total, average, minimum, or maximum number of times the tag was assigned to dialogs over the specified period will be counted.
-
Click Create report.
-
To reuse the report, click Save report. Specify its name and, optionally, a description.
The report will be available in the selected project in the Reports tab.
Applying a semantic attribute in a report
A semantic attribute is the dialog's feature that sets the context for analysis based on a search query. SpeechSense analyzes the dialog to see how close it is semantically to the search query in the context of the selected semantic attribute.
For example, you select the Themes semantic attribute and specify
Vacation in Sochias your search query. SpeechSense will analyze your dialogs to see whether or not their topic is semantically aligned withVacation in Sochi.
You can apply a semantic attribute in your report:
-
As a filter for dialogs (for Evaluation form, Comparison, and Details reports). The report will include only dialogs either relevant or not relevant to the search query in the context of the selected semantic attribute.
For example, you are using the Themes semantic attribute and the
Vacation in Sochisearch query as a filter. The report will then include only dialogs whose topic either does or does not match theVacation in Sochisearch query, based on your choice. -
As an evaluation parameter (only for Evaluation form). The semantic attribute will be one of the criteria used to evaluate the agent's performance.
For example, you are using the Results semantic attribute and the
Customer has booked a toursearch query as one of the evaluation parameters. In which case the agent's performance will be evaluated by whether or not the outcome of the dialog matches theCustomer has booked a tourquery.You can use a semantic attribute only once in each Evaluation form: either as a parameter or as a filter.
To apply a semantic attribute to dialogs in reports:
-
In the new report window, click
next to the Period field. Select Semantic attribute from the drop-down list. -
In the Semantic attribute setting window that opens, specify the period to pick the dialogs for to configure the semantic attribute. Add filters as needed. As a result, you will see the number of dialogs in the selection.
SpeechSense will analyze the specified selection of dialogs and generate suggestions for the specified semantic attribute. This requires at least 100 dialogs; otherwise, SpeechSense will not be able to conduct a reliable analysis. There are no restrictions on the maximum number; however, if the selection contains more than 1,000 dialogs, SpeechSense will randomly select only 1,000 dialogs for analysis.
If there is not enough or too much data in the selection, adjust the period and your filters.
Note
The period and filters you set in the Semantic attribute setting window are only used to select dialogs for semantic attribute configuration and do not affect the filtering of dialogs for the report.
-
Select a semantic attribute under YandexGPT criteria.
-
Put the cursor to the right of the semantic attribute in the same line and wait for suggestions to appear.
SpeechSense is now performing semantic analysis of selected dialogs and generating suggestions for the specified semantic attribute. If there are more than 1,000 dialogs in the selection and you are not satisfied with the suggestions that were generated, click Refresh for SpeechSense to re-select 1,000 dialogs at random and generate new suggestions.
-
Select a prompt or input your own.
-
Click Search.
SpeechSense will now analyze the dialogs again and display the results as two columns representing dialogs that do and do not conform to the specified conditions.
If not satisfied with the suggested dialog grouping, change your semantic attribute or search query. The dialogs will then be regrouped. You may need to change your criterion or query a number of times.
-
Once you are satisfied with the grouping, specify which dialog group should be used.
For example, if your criterion is Themes, click one of the two buttons:
- Contain the theme to use only dialogs with the specified topic.
- Do not contain theme to use dialogs with any topic other than the specified one.
Note
The dialogs presented in the columns will not necessarily match those included in the report, as different periods or filters may have been used to set up the semantic attribute and generate the report.
You can now add more filters and parameters and build your report.
-
In the new report window, click
Add parameter. Select Semantic attribute from the drop-down list. -
In the Semantic attribute setting window that opens, specify the period to pick the dialogs for to configure the semantic attribute. Add filters as needed. As a result, you will see the number of dialogs in the selection.
SpeechSense will analyze the specified selection of dialogs and generate suggestions for the specified semantic attribute. This requires at least 100 dialogs; otherwise, SpeechSense will not be able to conduct a reliable analysis. There are no restrictions on the maximum number; however, if the selection contains more than 1,000 dialogs, SpeechSense will randomly select only 1,000 dialogs for analysis.
If there is not enough or too much data in the selection, adjust the period and your filters.
Note
The period and filters you set in the Semantic attribute setting window are only used to select dialogs for semantic attribute configuration and do not affect the filtering of dialogs for the report.
-
Select a semantic attribute under YandexGPT criteria.
-
Put the cursor to the right of the semantic attribute in the same line and wait for suggestions to appear.
SpeechSense is now performing semantic analysis of selected dialogs and generating suggestions for the specified semantic attribute. If there are more than 1,000 dialogs in the selection and you are not satisfied with the suggestions that were generated, click Refresh for SpeechSense to re-select 1,000 dialogs at random and generate new suggestions.
-
Select a prompt or input your own.
-
Click Search.
SpeechSense will now analyze the dialogs again and display the results as two columns representing dialogs that do and do not conform to the specified conditions.
If not satisfied with the suggested dialog grouping, change your semantic attribute or search query. The dialogs will then be regrouped. You may need to change your criterion or query a number of times.
-
Once you are satisfied with the grouping, specify which dialog group should be used.
For example, if your criterion is Themes, click one of the two buttons:
- Contain the theme to use the match between the dialog and the given topic as an evaluation criterion.
- Do not contain theme to use the mismatch between the dialog and the given topic as an evaluation criterion.
Note
The dialogs presented in the columns will not necessarily match those included in the report, as different periods or filters may have been used to set up the semantic attribute and generate the report.
You can now add more parameters and filters and build your report.
Using semantic tags and semantic Pro tags for dialog evaluation
Semantic tags can help you evaluate the agent's performance and the client's behavior, e.g., whether the agent was polite, whether the client was rude, etc. To do this:
-
When creating a tag, select the appropriate semantic attribute and specify the relevant search query.
For example, if you are interested in dialogs where the customer was rude to the agent, select:
- Semantic attribute: Common question.
- Search query: Did the customer show rudeness or aggression in the conversation at least once?
-
Use a semantic tag as a report parameter.
Semantic Pro tags can help you evaluate the agent's performance and the client's behavior against complex semantic criteria, e.g., whether the agent was able to sell a travel package without offering a discount to the customer. To do this:
-
When creating one, specify a relevant search query.
For example, if you are looking for dialogs where a client bought a tour to the Maldives with no discount offered, specify the search query as follows: tour to the Maldives purchased without a discount.
-
Use the semantic Pro tag as a report parameter.