[Feat] Design Project Architecture

[pingpingy1]

Is your feature request related to a problem? Please describe. I should decide on the following things:

• [x] What kind of computational jobs are required?
• [x] Should I use Python or Prolog for each part?
• [x] How would the data flow between each module occur?
• [x] What non-computational functionalities are required? eg) User interaction

Describe the solution you'd like The resulting architecture must be modular, concise, and generalizable. For example, transitions from hard-coded phonetic similarity to IPA to vector DB should be smooth. Each module should have a well-defined, simple task that have little overlap. By looking at the architecture, the prototype should seem within reach. This should also lead to some loose (preferably tight) definition of the scope of functionalities.

Describe alternatives you've considered A totally-Python or totally-Prolog is impractical since a friendly interface as well as logical inferences are required. It may be beneficial to use some diagram-drawing software for this task. eg) Miro

Additional context

[pingpingy1]

Required computational tasks:

• Sentence parsing: Take the provided sentence and analyze its syntactic structure.
• Ambiguity localization: Determine the component to be replaced. If the sentence contains grammatical errors, this should be as simple(?) as fault localization; else, the problem is probably harder...
• Replacement searching: Once the ambiguity is localized, search for words/phrases that both satisfy grammatical, contextual, and phonetic requirements. Each requirement may be implemented as different modules.
• User interface: Take user's input of a joke and returns the analysis result (This is/is not a joke. This phrase, related to this phrase, should be replaced with this phrase, related to this phrase.)

Parsing, localization, replacement searching would be done with Prolog, while choosing a single replacement and interfacing with user would probably be done with Python.

[pingpingy1]

Miro link Here's a basic data flow diagram of the project. Given one, grammatically incorrect reading of the joke, its syntax tree is passed through the modules explained above to determine whether the sentence has a humorous homophone reading. The English, phonetic, and contextual KBs are designed to be separated from this process, so as to be as modular as possible.