About us

This website started in 2016 july 8th. The idea was to create a chatbot which could emulate a human. We started by making a simple sentence-sentence match association. This first version was made using text data in text files. We first wanted to make an A.I. but we knew it's more difficult than a simple chatbot. In the beginning we tried to find the best way of making a good association. Using text files was not the best way. For example we needed to code how to read the text file and how to modify a line in the text file. Even using text marker to lookup the file was complicated. We started to use database. With a database, we could handle the rows with sessions. It was a code-saving option. The main problem was to find a good way to retrieve the data and precisely but to keep allowing wide possibilities. We knew using sentence-sentence association wasn't the ideal way. Because there would be too many words and words like 'the' can't find a precise match. We needed to make distinction between the word. The only way to do this was to download a dictionnary which would tell if the word is: a verb, a nouns, an ajective, etc.. With the dictionnary we could analyse the sentence and store some keywords instead of a sentence. Now the goal was to recreate a different sentence using the user's inputed sentence. We had to take consideration of all the possibilities and all the grammar rules. We made a simple syntax builder for sentences. The sentence would pass in this filter and be reformulated as a question, a suggestion or an affirmation. By moving the words we could create new sentences. Now that the chatbot basics was made we needed to work on applying all the grammars rules. We also needed to find the reverse match for 'I' to 'You'. We had to detect if the user sentence is a question to respond with an affirmation instead of another question. We needed to know the position of each word in a sentence because a some words in the dictionnary are using the same wording. Therefore it was difficult to match the real meaning of the word to the dictionnary. We decided to make it learn by the question it's asking and by the use of possessive adjective like 'your'. Working with keywords, we had the idea to collect them in a session during the conversation to get more match from the database later on. We would collect every 4 answers some keywords that would help retrieve new rows from the chatbot memory. The next goal is to make it correct it's own mistakes. This would define it more as an A.I..