Overview

SUMO Natural Language Processing (SUMO NLP) takes natural language text and turns the text into logical statements. These statements can be queried against, and tested for inconsistencies.

Install pre-requisites

SUMO NLP is reliant on the the following:

• SUMO and SIGMAKEE. Installation instructions are here.
• Vampire. Installation instructions are here. Configuration instructions below.
• Miniconda (instructions below)
• Ollama (instructions below)

Create folder structure

Create a sumonlp folder where models and large files will be stored. This should be separate from where you clone this repository. For example:

cd ~
mkdir .sumonlp
cd .sumonlp
mkdir L2L_model

Add the following lines to your .bashrc file. HAMMING is a super computer at the Naval Postgraduate School. If you are not running on the Hamming super computer, this should be set to false:

export SUMO_NLP_RUNNING_ON_HAMMING=false
export SUMO_NLP_HOME=~/.sumonlp

Clone the repository to your workspace directory:

cd ~/workspace
git clone https://github.com/ontologyportal/sumonlp.git

Miniconda

Follow install instructions at https://docs.anaconda.com/miniconda/install/ . Linux instructions reproduced here for convenience:

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm ~/miniconda3/miniconda.sh
source ~/miniconda3/bin/activate
conda init --all

Create a conda environment. This allows you to create, export, list, remove, and update environments that have different versions of Python and/or packages installed in them.

As an example: If your version of Python were 3.9, then you would run:

conda create -n name_you_choose_for_environment python=3.9

Add the name of your environment ("name_you_choose_for_environment") to the .bashrc file:

export SUMO_NLP_CONDA_ENVIRONMENT="name_you_chose_for_environment"

Vampire Prover

Installation instructions are here. It is recommended to install vampire in After installing Vampire, add the following line to the .bashrc file:

export PATH="$HOME/path/to/vampire:$PATH" 

Configuration

The language to logic translator produces the file SUMO_NLP.kif, with logic statements generated from natural language. This file is automatically copied to the .sigmakee/KBs directory. To be able to translate the generated SUO-KIF logic to the TPTP input required by the vampire prover, update

$HOME/.sigmakee/KBs/config.xml

and add the following line is in the section:

...
  <kb name="SUMO" >
    ...
    <constituent filename="/home/THE_USER/.sigmakee/KBs/SUMO_NLP.kif" />
    ...
  </kb>
...

The files within this tag are all combined and translated into the SUMO.fof file which is then fed as input to the vampire prover.

Ollama Install Notes

Instructions adapted from: https://github.com/ollama/ollama/blob/main/docs/linux.md https://ollama.com/download/linux

cd Programs
mkdir ollama
cd ollama
curl -L https://ollama.com/download/ollama-linux-amd64.tgz -o ollama-linux-amd64.tgz
sudo tar -C ./ -xzf ollama-linux-amd64.tgz
cd bin
sudo chmod 777 *

To start a server, from the ollama/bin directory

./ollama serve

On different terminal, from the ollama/bin directory

./ollama -v

If you see a version, then it worked!

To run a specific model on the server (and download if necessary). For example:

./ollama run llama3.2 

For a list of models - https://ollama.com/library

Add the path of your ollama installation to the .bashrc file:

export OLLAMA_HOST_PORT="11434" # Used to change default port (11434) to unique port number if necessary.
export OLLAMA_HOST="127.0.0.1:$OLLAMA_HOST_PORT"
export PATH="$HOME/path/to/ollama/bin:$PATH" # Path to ollama executable

L2L model and vocabulary.db

The vocabulary.db file must be present, as well as the model used for language to logic conversion.

For more information, see here

Place the vocabulary.db file in $SUMO_NLP_HOME location, and model used for L2L conversion to $SUMO_NLP_HOME/L2L_Model.

Running sumonlp

1. To configure the models that will be run during different parts of the pipeline, edit the src/load_configs.sh file.
2. Run src/utils/install_requirements.sh. This will install necessary python packages.
3. Run src/main.sh
4. Type "help" to list available commands.

Running on Hamming GPU Node:

From a submit node, run

srun --pty -N 1 --partition=genai --gres=gpu:1 bash