Page 175 - AI Computer 10

The various steps involved in designing the model are:

Step 1: Identify User Intents
Some possible user intents might include:
u Inquiring about product information u Asking about return policy
u Checking order status u Requesting support contact details

Step 2: Define Rules
Some sample rules for the chatbot could be:

u Rule 1: If user input contains the words "product" or "info", then respond with "Which product do you wish
to know about?"
u Rule 2: If user input contains the words "order" or "status", then respond with "Please provide your order
number to know more."

u Rule 3: If user input contains the words "return" or "refund", then respond with "Our return policy allows
returns within 30 days. Would you like to know how to initiate a return?"
u Rule 4: If user input contains the words "contact" or “support” or “executive”, then respond with "You can
reach our customer support at support@example.com or call us at 1880-1800-189”.

u Rule 5: If no match found, respond with "I'm sorry, I didn't understand that. Can you please rephrase your
question?"
Step 3: Interaction
When a user interacts with the chatbot, their message is examined according to the established rules. The chatbot
then replies with a predefined answer or asks the user for further information, depending on the situation.
Rule-based AI models are easy to interpret and implement. They are useful where the set of conditions or
circumstances in which the model has to operate are limited, clearly defined, and have a well-defined set of
responses that most receivers will understand. These models require minimal computing power as no complex
computations are involved.
A major drawback of rule-based system is that the learning process is static. Once the machine is trained, it
does not account for any changes made to the original training dataset. If you test the machine on a dataset
that differs from the rules and data used during training, it will likely fail. Another limitation of the rule-based
model is that the model does not learn from its mistakes. After training, the model is unable to adapt based on
feedback from the users. Updating the model by including new rules can be a time-consuming process.

Learning Based Approach
Learning Based Approach, or Adaptive Intelligence Approach, refers to an AI model where relationship or patterns
in the data are not defined by the developer, rather the model learns on its own using the input data and the
feedback that its actions generate. A learning-based approach system does not rely on static data as random
data is fed into the machine and it generates output on the basis of its own identification of patterns or trends.
Remember that rule-based systems are static, whereas learning based systems are dynamic due to the adaptability
of new user inputs and information. As the system gets updated with new data, frequent and regular training of
the model is required in learning-based systems, which is one of the major drawbacks of these systems.

Examples of Learning Based AI Models
Example 1: Suppose you have a dataset of 1000 images of butterflies, but you do not have any clue as to what
trend is being followed as you don’t know their species, colour, size, or any other feature. Thus, you would put

41
41