Refactor to connect with the rest of the modules.

• Move strings from CT to the specific folder
• Increase the options to have more variety of dialogues
• Add examples for: interesting places, restaurant search, goodbye, open question

A) We should create submodules for separate dialogue acts (specific Classes):

• restaurant search
• interesting places (we can use 2nd Notebook I sent to search with Wikipedia API)
• google maps integration (for directions)
• weather??? (later)

B) Types of interactions: utterances, intentions, entities (slots)

• restaurant search for particular cuisine
• places of interest (by direct question or by open question)
• weather??? (later)

This feels good!

A) We should create submodules for separate dialogue acts (specific Classes):

• restaurant search
• interesting places (we can use 2nd Notebook I sent to search with Wikipedia API)
• google maps integration (for directions)
• weather??? (later)

B) Types of interactions: utterances, intentions, entities (slots)

• restaurant search for particular cuisine
• places of interest (by direct question or by open question)
• weather??? (later)

Check GDrive doc

Check Gdrive doc

Now this projects contains 3 initial proposed scripting that work following the modules idea found on the previous link.

Inside folder imutils:

• dmanager.py --> Dialogue manager. Contains a class with dialogue functions. From here we move to all other scripting

• processing_layer.py --> Audio Input to text Output processing scripting

• semantic_layer.py --> Context, History and connection to DB (maybe we can split this)

Inside folder media:

• Here, all audios are saved

Inside folder templates:

• Just the html

Inside folder doc:

• Requirement doc and all other coding-related doc

Inside folder notebooks:

• Probably you like to test using notebooks and also it is good to explain how things work. Now there is only a notebook about pre-processing and classification tasks.

Now the project is made using Flask so there is also a app.py script. We will only keep this if we decide to use Flask.

A) We should create submodules for separate dialogue acts (specific Classes):

• restaurant search
• interesting places (we can use 2nd Notebook I sent to search with Wikipedia API)
• google maps integration (for directions)
• weather??? (later)

B) Types of interactions: utterances, intentions, entities (slots)

• restaurant search for particular cuisine
• places of interest (by direct question or by open question)
• weather??? (later)

As the title says, it is not as good as we would like it to be. The main problem at least in my computer is that it sometimes fails to reproduce the audio. This is solved using a try-catch but it makes the system slow, waiting it gets up again.

• Look for other SR systems
• Try to improve this one (some quick research)
• Maybe increase to 10sec for the user to speak

• PoS Tagging
• Dependency Parser

We can use the example Natu sent (first Notebook), which uses spaCy.

TODO:

• Create the model (improve training) for the dialog examples we have on the Doc
• From the inputs, by using spacy, you select main words depending on the utterance detected

A) We should create submodules for separate dialogue acts (specific Classes):

• restaurant search
• interesting places (we can use 2nd Notebook I sent to search with Wikipedia API)
• google maps integration (for directions)
• weather??? (later)

B) Types of interactions: utterances, intentions, entities (slots)

• restaurant search for particular cuisine
• places of interest (by direct question or by open question)
• weather??? (later)