Code to create an AI to play the game Codenames as a Spymaster.
The idea is that the spy master must think of a word (that is not on the game) which connects the most amount of words from his team (blue or red) without connecting to the other team's or black cards.
If he does a good job at it, the Operatives (other players) will have an easier time trying to guess which cards are from their team and which are not.
This is the game interface:
If the spymaster was on the red team, saying the word "voador" (flyer) might be a good call to connect the words "drone" and "para-quedas" (parachute) for the operators to choose them. On the other hand, the blue spymaster could make the mistake of saying "animal" trying to connect the words "leão" (lion) and "peru" (turkey) forgetting that the red team has the card "elefante" (elephant) and the operatives won't have any way of knowing which one is correct.
First you need to generate the word embeddings using the generateencodings.py script like
python generateencodings.py --model models/ptwiki_20180420_100d.txt --words data/words.txt --output embeddings_100d.txt
That will create the embeddings_100d.txt file that contains the embeddings for all words contained in data/words.txt using the model models/ptwiki_20180420_100d.txt.
Then, with the embeddings file created, we can run the script with
python main.py --team red --encodings embeddings_100d.txt