(Modified comment string {{Template:PEC_Migrated}} with __NOINDEX__ {{Template:PEC_Migrated}}) |
|||
| Line 6: | Line 6: | ||
| − | The Trending page is designed to help you monitor and analyze terms and phrases within conversations and identify and explore familiar and unexpected trends. By analyzing term trends within a specific interaction set | + | The Trending page is designed to help you monitor and analyze terms and phrases within conversations and identify and explore familiar and unexpected trends. By analyzing term trends within a specific interaction set over specific time periods and/or according to a variety of search criteria (for example, agent and/or customer side, metadata, duration, categories, topics, agents, work groups, language and so on), Trending enables you to better understand emerging business issues, pinpoint events that may require close attention, identify process or service issues before they escalate and recognize strengths and weaknesses of the organization’s employees, products, and processes. |
| − | The data in the Trending | + | The data in the Trending page is based on an automated analysis of interaction transcripts together with the identification of changes in the frequency with which a term/phrase appears in interactions. Term/phrase frequency is the number of interactions within which the term/phrase is found divided by the entire set of interactions during a given time period, expressed as a percentage. For example, if during a specific time period a phrase was used in 350 out of 1000 interactions; the phrase frequency for that given time period is (350/1000)*100% = 35%. |
| − | Terms/phrases appear in the Trending chart according to a | + | Terms/phrases appear in the Trending chart according to a salience measurement (for example, relative importance measurement). Terms and phrases with a high salience measurement will be included in the Trending chart. The purpose of the salience measurement is to filter out common terms/phrases (for example, thank you), to make room for significant terms and phrases. That is, a phrase is more salient if it does not appear in too many interactions. For example, a phrase that appears in 10-15% of the interactions will more likely have a higher saliency then a phrase that appears in 30% of the interactions. |
| − | Trending requires no input from users about | + | Trending requires no input from users about what terms and phrases should be found within the interactions and for this reason the information in the Trending page can help you identify issues and trends that are unforeseen or unexpected. That is, Trending is a discovery feature, not a search feature. It’s not about searching for a specific phrase. Instead, it’s about exploring a specific time period, speakers, and additional metadata. |
| − | From the Trending page you can create and save Trending charts that show the distribution of terms and phrases within a selected interaction set. The relative frequency of terms and phrases is visualized using a bubble chart. The bubble chart (see image below) provides at-a-glance insight on four dimensions: | + | From the Trending page you can create and save Trending charts that show the distribution of terms and phrases within a selected interaction set or you can create and save charts that compare the frequency between two sets of interactions. The relative frequency of terms and phrases is visualized using a bubble chart and shown in a table. The bubble chart (see image below) provides at-a-glance insight on four dimensions: |
| − | * '''Size: '''The size of the bubble is proportionate to the number of times the term or phrase appears in the interaction set. The larger the bubble’s radius, the more the term or phrase is used within the interaction set. | + | * '''Size: '''The size of the bubble is proportionate to the number of times the term or phrase appears in the interaction set. The larger the bubble’s radius, the more the term or phrase is used within the interaction set. |
| − | * '''Color:''' The bubbles are shaded red or green and represent the change in frequency. The deeper the shading the larger the frequency change. Red represents a negative (reduction) in frequency and green represents a positive (increase) in frequency for that particular term. | + | * '''Color:''' The bubbles are shaded red or green and represent the change in frequency. The deeper the shading the larger the frequency change. Red represents a negative (reduction) in frequency and green represents a positive (increase) in frequency for that particular term. |
* '''Comparison:''' Bubbles that contain a dotted line circle indicate a term or phrase that appear in the selected interaction set in both Period 1 and Period 2. For example, if you are viewing the Period 1 tab, the dotted line represents the term/phrase frequency in Period 2 and vice versa. | * '''Comparison:''' Bubbles that contain a dotted line circle indicate a term or phrase that appear in the selected interaction set in both Period 1 and Period 2. For example, if you are viewing the Period 1 tab, the dotted line represents the term/phrase frequency in Period 2 and vice versa. | ||
| − | * '''Cluster:''' Bubbles connected with lines represent a collection | + | * '''Cluster:''' Bubbles connected with lines represent a collection of terms and phrases within the interaction set that have a strong semantic similarity. |
| Line 27: | Line 27: | ||
This visual representation enables you to quickly identify the terms and phrases with the highest and lowest frequency, and any deviation from regular term frequencies. This enables users to surface and analyze trends that might otherwise go undetected. After recognizing the terms and phrases that have the strongest or weakest correlation, you can refine your search results accordingly, to search for the interactions most related to the business issue you are investigating. | This visual representation enables you to quickly identify the terms and phrases with the highest and lowest frequency, and any deviation from regular term frequencies. This enables users to surface and analyze trends that might otherwise go undetected. After recognizing the terms and phrases that have the strongest or weakest correlation, you can refine your search results accordingly, to search for the interactions most related to the business issue you are investigating. | ||
| + | |||
| + | The Trending backend mechanism is comprised of two stages: | ||
| + | |||
| + | '''Offline (cluster task) process'''<br> | ||
| + | The offline process periodically runs a clustering task on a set of interactions that are defined using a search filter. The result is a set of phrases that will be used during the online process. The phrases resulting from the clustering task are sorted by saliency. | ||
| + | |||
| + | '''Online (filter task) process'''<br> | ||
| + | The online process occurs in the SpeechMiner UI. A search for the phrases identified in the offline task is performed on the interactions that match a specific filter criterion. The matching interaction count for each phrase is displayed in the UI. | ||
For detailed information about how to configure and interpret the Trending page, refer to the following: | For detailed information about how to configure and interpret the Trending page, refer to the following: | ||
| Line 36: | Line 44: | ||
* [[blacklist|Manage the Blacklist]] | * [[blacklist|Manage the Blacklist]] | ||
* [[savedtrendingfilters|Working with Saved Trending Filters]] | * [[savedtrendingfilters|Working with Saved Trending Filters]] | ||
| + | |||
[[Category:V:PSAAS:Public]] | [[Category:V:PSAAS:Public]] | ||
Revision as of 09:38, October 11, 2020
Trending
The Trending page is designed to help you monitor and analyze terms and phrases within conversations and identify and explore familiar and unexpected trends. By analyzing term trends within a specific interaction set over specific time periods and/or according to a variety of search criteria (for example, agent and/or customer side, metadata, duration, categories, topics, agents, work groups, language and so on), Trending enables you to better understand emerging business issues, pinpoint events that may require close attention, identify process or service issues before they escalate and recognize strengths and weaknesses of the organization’s employees, products, and processes.
The data in the Trending page is based on an automated analysis of interaction transcripts together with the identification of changes in the frequency with which a term/phrase appears in interactions. Term/phrase frequency is the number of interactions within which the term/phrase is found divided by the entire set of interactions during a given time period, expressed as a percentage. For example, if during a specific time period a phrase was used in 350 out of 1000 interactions; the phrase frequency for that given time period is (350/1000)*100% = 35%.
Terms/phrases appear in the Trending chart according to a salience measurement (for example, relative importance measurement). Terms and phrases with a high salience measurement will be included in the Trending chart. The purpose of the salience measurement is to filter out common terms/phrases (for example, thank you), to make room for significant terms and phrases. That is, a phrase is more salient if it does not appear in too many interactions. For example, a phrase that appears in 10-15% of the interactions will more likely have a higher saliency then a phrase that appears in 30% of the interactions.
Trending requires no input from users about what terms and phrases should be found within the interactions and for this reason the information in the Trending page can help you identify issues and trends that are unforeseen or unexpected. That is, Trending is a discovery feature, not a search feature. It’s not about searching for a specific phrase. Instead, it’s about exploring a specific time period, speakers, and additional metadata.
From the Trending page you can create and save Trending charts that show the distribution of terms and phrases within a selected interaction set or you can create and save charts that compare the frequency between two sets of interactions. The relative frequency of terms and phrases is visualized using a bubble chart and shown in a table. The bubble chart (see image below) provides at-a-glance insight on four dimensions:
- Size: The size of the bubble is proportionate to the number of times the term or phrase appears in the interaction set. The larger the bubble’s radius, the more the term or phrase is used within the interaction set.
- Color: The bubbles are shaded red or green and represent the change in frequency. The deeper the shading the larger the frequency change. Red represents a negative (reduction) in frequency and green represents a positive (increase) in frequency for that particular term.
- Comparison: Bubbles that contain a dotted line circle indicate a term or phrase that appear in the selected interaction set in both Period 1 and Period 2. For example, if you are viewing the Period 1 tab, the dotted line represents the term/phrase frequency in Period 2 and vice versa.
- Cluster: Bubbles connected with lines represent a collection of terms and phrases within the interaction set that have a strong semantic similarity.
Bubble Chart Example
This visual representation enables you to quickly identify the terms and phrases with the highest and lowest frequency, and any deviation from regular term frequencies. This enables users to surface and analyze trends that might otherwise go undetected. After recognizing the terms and phrases that have the strongest or weakest correlation, you can refine your search results accordingly, to search for the interactions most related to the business issue you are investigating.
The Trending backend mechanism is comprised of two stages:
Offline (cluster task) process
The offline process periodically runs a clustering task on a set of interactions that are defined using a search filter. The result is a set of phrases that will be used during the online process. The phrases resulting from the clustering task are sorted by saliency.
Online (filter task) process
The online process occurs in the SpeechMiner UI. A search for the phrases identified in the offline task is performed on the interactions that match a specific filter criterion. The matching interaction count for each phrase is displayed in the UI.
For detailed information about how to configure and interpret the Trending page, refer to the following: