A Novel Framework for Arabic Dialect Chatbot Using Machine Learning

Table 5

Comparison of the chatbots (our chatbot and Pandorabot).

Comparison criteria	Chatbot produces by using our proposal framework	Chatbot produces by using Pandorabot platform
Type of dataset	- Use unstructured Arabic data (text/yaml). And there is no need to define each question/answer because our chatbot can search through all corpus within seconds to retrieve the data.	- Use AMIL file required to define each question as <pattern></pattern> and the answer as <template></template> for each dialog.; - The question must be in capital letter (English case).; - The data must not have any kind of punctuations because the platform will ignore it immediately and will not deploy it.
Time of response	The answer was retrieved immediately.	The answer took around 30 seconds to retrieve, and it could be “apologize message.”
Dealing with user input	- The user can insert any Arabic input by using the formal and informal Arabic language.	- The user can insert only the exact question already fixed in the AIML file. Except that, the chatbot will ignore the question and retrieve “apologize message.”
The experimentation	We insert the same dataset for both bots. And we use the same question format. The question used is “نسيت كلمة مرور جوالي الايفون” and the answer should be; “نسيت كلمة مرور” ابل أي دي “في الايفون:الإعدادات>انقر على [اسمك] > كلمة السر والأمن > تغيير كلمة السر” which contained diagnosis of the problem followed by the solution.
The result	- If the question is in informal Arabic language, the bot retrieves the correct answer. Input: “نسيت كلمة مرور جوالي الايفون”; Output: the bot retrieves the correct answer.	- If the question is in informal Arabic language, the bot retrieves “apologize message.”; Input: “نسيت كلمة مرور جوالي الايفون”; Output: the bot retrieves error “apologize message.”; - If the question is in an exact AIML file premade, the bot retrieves the correct answer.; Input: نسيت كلمة مرور “ابل أي دي” في الايفون; Output: the bot retrieves the correct answer.
Finding	- Our chatbot was capable of having any type of Arabic data without having strict inputs to a specific form. The bot uses AI keyword matching to analyze user input and match it with the most relevant problem/solution.	- Pandorasbot working as simple chatbot was capable of matching a text string and offering an answer only when the exact sentence match is found.