This repository is provided for general use and as the basis for the assignment for Computational Logic for Artificial Intelligence (COMSM0022). CDT students taking the assignment should clone this repository. The two buttons above give more details about the assignment, and enable running the Prolexa assistant in the browser using Google Colab. The rest of this page provides a bit more details about what the code can do, how it can be run from the command line, and how it can be integrated with an Amazon Alexa device.
This repository contains Prolog code for a simple question-answering assistant.
The top-level module is prolexa/prolog/prolexa.pl, which can either be run in
the command line or with speech input and output through the
alexa developer console.
The heavy lifting is done in prolexa/prolog/prolexa_grammar.pl, which defines
DCG rules for sentences (that are added to the knowledge base if they don't
already follow), questions (that are answered if possible), and commands (e.g.,
explain why something follows); and prolexa/prolog/prolexa_engine.pl, which
implements reasoning by means of meta-interpreters.
Also included are prolexa/prolog/nl_shell.pl, which is taken verbatim from
Chapter 7 of Simply Logical, and an extended version
prolexa/prolog/nl_shell2.pl, which formed the basis for the prolexa code.
The code has been tested with SWI Prolog versions 7.6.0, 8.0.3 and 8.2.2.
(The code is executed from the prolexa/prolog directory.)
% swipl prolexa.pl
Welcome to SWI-Prolog (threaded, 64 bits, version 8.0.3)
SWI-Prolog comes with ABSOLUTELY NO WARRANTY. This is free software.
Please run ?- license. for legal details.
For online help and background, visit http://www.swi-prolog.org
For built-in help, use ?- help(Topic). or ?- apropos(Word).
?- prolexa_cli.
prolexa> "Tell me everything you know".
*** utterance(Tell me everything you know)
*** goal(all_rules(_7210))
*** answer(every human is mortal. peter is human)
every human is mortal. peter is human
prolexa> "Peter is mortal".
*** utterance(Peter is mortal)
*** rule([(mortal(peter):-true)])
*** answer(I already knew that Peter is mortal)
I already knew that Peter is mortal
prolexa> "Explain why Peter is mortal".
*** utterance(Explain why Peter is mortal)
*** goal(explain_question(mortal(peter),_8846,_8834))
*** answer(peter is human; every human is mortal; therefore peter is mortal)
peter is human; every human is mortal; therefore peter is mortal
Follow the steps below if you want to use the Amazon Alexa speech to text and text to speech facilities. This requires an HTTP interface that is exposed to the web, for which we use Heroku.
swipl -g "mk_prolexa_intents, halt." prolexa.pl
The intents are found in prolexa_intents.json. You can copy and paste the
contents of this file while building your skill on the
alexa developer console.
To build:
docker build . -t prolexa
To run:
docker run -it -p 4000:4000 prolexa
To test the server:
curl -v POST http://localhost:4000/prolexa -d @testjson --header "Content-Type: application/json"
Prerequisites:
- Docker app running in the background.
- Installed Heroku CLI (
brew install heroku/brew/herokuon MacOS).
To see the status of your Heroku webapp use
heroku logs
in the prolexa directory.
-
Clone this repository
git clone [email protected]:So-Cool/prolexa.git cd prolexa -
Login to Heroku
heroku login -
Add Heroku remote
heroku git:remote -a prolexa
-
Before you start open your local copy of Prolexa and login to Heroku
cd prolexa heroku container:login -
Change local files to your liking
-
Once you're done push them to Heroku
heroku container:push web heroku container:release web -
Test your skill and repeat steps 2. and 3. if necessary
-
Once you're done commit all the changes and push them to GitHub
git commit -am "My commit message" git push origin master
Prolexa Plus is an extension to Prolexa which uses NLTK and Flair for part-of-speech tagging of nouns, verbs and other words that are not currently in Prolexa's lexicon. It was implemented by Gavin Leech and Dan Whettam from the CDT19 cohort.
% python prolexa/prolexa_plus.py
2020-11-10 18:33:12,559 loading file /Users/cspaf/.flair/models/en-pos-ontonotes-v0.5.pt
Hello! I'm ProlexaPlus! Tell me anything, ask me anything.
> tell me about Kacper
*** utterance(tell me about Kacper)
*** goal(all_answers(kacper,_60700))
*** answer(I know nothing about kacper)
I know nothing about kacper
> Kacper is a postdoc
*** utterance(Kacper is a postdoc)
*** rule([(postdoc(kacper):-true)])
*** answer(I will remember that Kacper is a postdoc)
I will remember that Kacper is a postdoc
> every postdoc is busy
*** utterance(every postdoc is busy)
*** rule([(busy(_53392):-postdoc(_53392))])
*** answer(I will remember that every postdoc is busy)
I will remember that every postdoc is busy
> Kacper is busy
*** utterance(Kacper is busy)
*** rule([(busy(kacper):-true)])
*** answer(I already knew that Kacper is busy)
I already knew that Kacper is busy
> explain why Kacper is busy
*** utterance(explain why Kacper is busy)
*** goal(explain_question(busy(kacper),_3046,_2824))
*** answer(kacper is a postdoc; every postdoc is busy; therefore kacper is busy)
kacper is a postdoc; every postdoc is busy; therefore kacper is busy
Prolexa Plus requires Python 3.6+ and SWI Prolog version 7.6.0+. Using a Python virtual environment is advised. Since the Prolog<->Python bridge is quite fragile, you should consider using:
- Windows Subsystem for Linux if you are on Windows,
- Ubuntu Linux,
- MacOS, or
- the provided Docker image (see below)
to minimise potential issues.
For more information on how to set up pyswip see:
This installation approach is recommended.
The installation script may take a moment when processing the Prolexa package
since language models need to be downloaded (which is achieved by automatically
executing the prolexa/setup_models.py script) -- the
Running setup.py install for prolexa ... / step.
To install execute
pip install -e .
while in the root directory of this repository.
The -e flag installs an editable version of the package, which allows you to
edit the source to instantly update the installed version of the package
(read more
here).
This installation comes with two command line tools:
prolexa-plus-- launches the Prolexa Plus CLI, andprolexa-setup-models-- downloadsnltkandflairlanguage corpora and models.
- Install Python dependencies
pip install -r requirements.txt - Install language models and data
python prolexa/setup_models.py - Run Prolexa Plus
PYTHONPATH=./ python prolexa/prolexa_plus.py
Instead of a local install, it is possible to run Prolexa Plus with the designated Docker image.
- Build the Prolexa Plus Docker image
docker build -t prolexa-plus -f Dockerfile-prolexa-plus ./ - Run Prolexa Plus via Docker
docker run -it prolexa-plus
Python tests are currently broken. To test the code execute
python prolexa/tests/test.py