Chatbot
Building Your Bot
Message History
10 min
message history / user chat history the admin can see the comprehensive message history and sentiments of all the users interacting with the chatbot the admin can export these logs through keyword or for a period an option to filter by queries, feedback, channels, sentiments, timestamp and location is provided the user can also toggle between message history and user history an option to filter by channel and date in the user chat history identified intent analytics analytics to determine how much our nlp was when the bot is responding to a user query the admin can see the comprehensive message history and sentiments of all the users interacting with the chatbot the admin can export these logs through keyword or for a period the user can also toggle between message history and user history message history filter this is visible under message history > toggle from user chat history to message history > click on apply filter type in queries these are the queries which are typed in manually or sent without without selecting any suggestions, recommendations, menu, sticky, child intents by the customers this is visible under cogno chatbot console > message history > toggle to message history > apply filter > type in queries identified intent analytics analytics to determine how much our nlp was when the bot is responding to a user query this is visible under chatbot console > message history new metrics have been added to the "message history" table variation responsible this indicates when an intent is triggered against a customer query inside the chatbot, which training sentence variation of that intent was having the nearest match with the customer query which led to that intent being triggered % match this indicates the percentage match of a particular intent (variation/keyword/intent name) with a customer query when that particular intent gets triggered average accuracy this indicates the average accuracy percentage of the bot's recommendation system and the same get's calculated as a mean of all the values getting displayed on the "%match" field inside message history at that instant % match formula \[( customer query segmented words (other than stop words) / nearest training sentence segmented words or intent name segmented words)] 100 user friendly ui/ux for the message history page columns can now be edited to keep the view of only the data the user wants to view this can be done by clicking on the "edit" icon available on the top right side of the message history page no of records per page by default would be 10 and the same can be changed to 25, 50, 70 and 100 from the "no of records per page" dropdown available just on the left side of the "filter" button two more filters to filter queries which are available inside the filter a dropdown to select intent when an intent is selected here only those queries will get loaded where that particular intent has been triggered by the bot against a customer query percentage bracket slider when a user selects a particular %bracket, only the customer queries where recognized intent % fall under the provided bracket would be displayed please note the "% match" and "variation responsible" data wouldn't get displayed by default to turn it on the user has to access a toggle button present in the backend currently when the user exports this table in xls format the 2 new data fields will also be available in the xls report this report would contain data for these 2 fields based on the timings these settings have been kept on from the backend for eg if users exports the xls report over a period of 1 week and within that 1 week, the toggle button to turn on these data fields were turned on for day 1 and day 3 and not for the other days, the report would display "variation responsible" and "%match" data only in day1 and day 3 of that week and not on other days queries can be sorted by percentage (high to low, low to high) by clicking on the %match column header they can also be sorted alphabetically by clicking on the "user query" column header the “average accuracy” data won't be displayed if the toggle button to turn this setting on is set to off from the backend average accuracy gets calculated based on the queries getting displayed at a particular inside message history having %match values (as a mean of this) if somebody has filtered the queries as "type in" queries the average accuracy value would be displayed based on the mean of "%match" for the type in queries getting displayed at that instant if necessary keywords, restricted keywords, and intent level threshold score are set for an intent, additional validations related to the above mentioned configurations would be needed, and even if the percentage match stays more than other intents with a user query, absence of one necessary keyword in the query or presence of one restricted keyword in the query would restrict that particular intent from being triggered "recognised intent" filters would only filter queries where a response has been provided by the bot and won't fetch queries where the bot just gave suggestions there wouldn't be a "variations responsible" and "% match" value for queries where the bot just gave suggestions tag as false positive providing an option for the users to to set a percentage threshold inside django admin to configure to set a percentage threshold inside django admin → "percentage threshold for message history" all user queries having a percentage match below this will be flagged queries and the percentage representation of those queries would turn red (default value = 25%) added a filter "flagged queries" using which users can filter out queries which are flagged (whose percentage match lies below the specified percentage inside django admin) also providing an option to tag "flagged queries" as "false positives" and "not false positives" this is visible under chatbot console > message history > toggle towards messsage history > select a user query to view options for false positive/not false positive also providing an analytical view of the number of false positives on a day to day basis inside "message analytics" card a toggle button is provided in django admin to enable the analytical view for false positives, the column view inside message history for false positive and not false positive status, and filter for false positives and not false positives please note by default it will be enabled also while exporting the sheet, these values will only reflect in the exported sheet if the toggle for the same is turned on in django admin ui updates queries with a percentage match below the specified percentage threshold in django admin will be denoted in red colour refer the below shown demo for better understanding please note (limitations) if a specific query is not tagged "false positive" or "not false positive", the value for the same query inside "tags" column in the exported sheet will be blank and will be represented with " " if the toggle to turn on false positives release is enabled, the toggle for variation responsible and percentage match will also be enabled by the user chat history messages can be filtered based upon channels and dates in user chat history this is visible under message history > apply filter > select daterange/channel(s) > click on apply filter on applying filter chats would be shown as per the selected parameters in filter unique sessions for the same user session ids will be maintained for multiple sessions created by a unique user a session id column in the message history table will redirect the console user to the particular session in the user chat history when the user clicks on the session id this is visible under chatbot console > analyse and improve > message history > user chat history > message history the remaining functionality remains the same as applicable for another column in the table metadata > message history i e add/remove a column, change the position of the column please note session id will be captured for all channels also, there is no change done w\ r t the session capture for the following channels web android ios whatsapp session id for all other channels for a specific user will be counted at eod eg for user id insta 7171828299, session will end at the eod (11 59 pm) clicking on the specific session id, the console user will be redirected to the user chat history page user chat history page this is visible under chatbot console > analyze and improve > message history > user chat history ui and ux improvements have been done on the user chat history page a new session indicator is added which displays the session id and user query count every time a new session is initiated, the indicator will show the bifurcation in the chat session id dropdown is added to switch between sessions seamlessly selecting a session id from the dropdown will scroll down to the selected session the search bar added in the user id header will help in searching for any particular message(s) of the user console users need to simply type in and hit enter to search for the content adding a user query to the intent’s training question can now simply be done by hovering over the particular query > click on (+) indicator > select using check bot > add to intent > select intent (refer to the above shown example for a better understanding of the same) business initiated & customer initiated bifurcation in message history
🤔
Have a question?
Our knowledgeable support team and an awesome community will get you an answer in a flash.
To ask a question or participate in discussions, you'll need to authenticate first.