Bots – Interpreting Customer Intention
At a Glance
- Leveraging the Integrated Natural Language Processing (NLP)
- Associating customer expressions with probable intentions
- Connecting intentions to corresponding response actions
Regarding Bots – Deciphering Customer Intent
Softdial Contact Center™ (SCC) offers a comprehensive platform for creating diverse text bots, ranging from straightforward dialogues with direct customer queries to more advanced conversations capable of interpreting complex customer sentences and responding in various ways.
Well-configured chatbots enhance the customer experience by saving time and improving contact center efficiency.
To provide top-notch service, chatbots must grasp the customer’s intention, regardless of how it’s expressed. To empower chatbots, SCC’s flexible flow design tool, Softdial Scripter™, provides:
- A rich toolkit for text processing.
- Text-based Natural Language Processing (NLP) capabilities.
This information is divided into two sections:
- Using NLP to Understand Customer Intention: Explores the use of NLP to comprehend customer intent effectively.
- Building an Intentions Directory: Covers the creation of a directory to categorize and manage customer intentions.
Part 1 – Using NLP to understand customer intention
Figure 1 presents a straightforward example using one of the options provided by Softdial Scripter’s toolset for building chatbots. All these functions are facilitated by accessing customer data and creating and utilizing specialized dictionaries available within Scripter.
There’s no definitive method for developing chatbot applications; the approach depends on the nature of the bot’s tasks and how the Bot owner intends to interact with customers. In practice, Scripter may generate several scenarios, and the one(s) with the most effective responses is/are typically selected.
In this particular scenario, an internal chatbot manages a TV package upgrade using Scripter’s text-based NLP engine.
In this flow example:
- A conversation is initiated by either the chatbot or the customer.
- The chatbot queries the customer, verifying their customer ID.
- Upon successful verification, the chatbot poses an open-ended question: “How can I assist you?”
- The customer requests an upgrade.
At this juncture, specifically in the “Answer” step, Scripter can employ NLP to decipher the customer’s intention.
Inside Natural Language Processing (NLP)
Scripter’s NLP engine divides the customer’s message into sentences and words, categorizing the words as follows:
This process generates a list of terms that can be cross-referenced with a knowledge base to ascertain the appropriate action, such as upgrading the TV package.
For instance, when a customer writes, “I would like to upgrade my TV package,” the analysis of this sentence yields a set of terms – like, upgrade, TV package – which likely represent the customer’s intention. These terms can be utilized to search a multi-term dictionary within the Scripter text processing library, helping determine if there is a matching action.
Suppose within our dictionary of actions, we maintain the following association:
After interpreting the customer’s intention, the corresponding action is to provide a selection of upgrade packages and inquire about the customer’s preference.
The script flow will persist in this loop until the final action, which could involve sending a message to the appropriate ERP system to proceed with the chosen package upgrade.
Part 2 – Building an intentions directory
To enable a chatbot to generate coherent responses to customer inquiries, it’s essential to have an intentions dictionary that establishes the relationship between pertinent terms and outlines potential intentions for the chatbot.
To construct this dictionary, the supervisor will utilize Scripter Designer to define the most relevant responses to various inbound text messages.
In Figure 2, you can observe how the Intentions Analyzer is employed to extract intentions from the words used by customers.
– Verbs like ‘like’ or ‘want’ are recognized and categorized as ‘Positive Intention.’
– Nouns like ‘TV’ or ‘package’ identify the specific items the text pertains to.
Once individual words in a phrase have been identified and tagged with specific intents, these words can be amalgamated to deduce the overall intent of the phrase.
For instance, if the NLP system recognizes words belonging to the ‘Positive Intentions,’ ‘TV Packages,’ and ‘Upgrade’ intents, the amalgamation of these three categories enables the NLP system to ascertain the overall intent of the phrase. In this case, it discerns that the customer intends to upgrade their TV package, as illustrated in Figure 3.
Catalog of intentions correlated with corresponding responses
Numerous combinations of intentions can exist, each signifying a distinct meaning and demanding a distinct response. To address this, it’s feasible to establish a repository of intentions and associate each intention with an appropriate response to be conveyed to the customer.
Furthermore, SCC’s Scripter engine can be activated to execute tasks beyond simply responding with text. For instance:
– Routing the session to a live agent equipped with the requisite skills to manage the session.
– Conducting a database lookup to retrieve an account balance and furnishing the outcome to the customer.
