juego_embbedings_text_config.py

# %%
import json
import pickle as pk
import random
import threading
from datetime import datetime

import numpy as np
from gensim.models import KeyedVectors
from sentence_transformers import SentenceTransformer

from display import display_words
from pistas import curiosity, hint
from seguimiento import calculate_moving_average, calculate_tendency_slope

# %%
model = KeyedVectors(768)
model_st = SentenceTransformer(
    "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
)
# file_path = "word_embeddings_mpnet.pth"
# embeddings_dict = torch.load(file_path)
embeddings_dict = {}

config_file_path = "config/lang.json"
secret_file_path = "config/secret.json"


class DictWrapper:
    def __init__(self, data_dict):
        self.__dict__.update(data_dict)


with open(config_file_path, "r") as file:
    # Load JSON from the file into a dictionary
    Config_full = json.load(file)

with open(secret_file_path, "r") as file:
    # Load JSON from the file into a dictionary
    secret = json.load(file)

lang = 0

if lang == 0:
    Config = DictWrapper(Config_full["SPA"]["Game"])  # type: ignore
    secret_dict = secret["SPA"]
elif lang == 1:
    Config = DictWrapper(Config_full["ENG"]["Game"])  # type: ignore
    secret_dict = secret["ENG"]
else:
    Config = DictWrapper(Config_full["SPA"]["Game"])  # type: ignore
    secret_dict = secret["SPA"]


with open("ranking.txt", "w+") as file:
    file.write("---------------------------")

# %%
pca = pk.load(open("pca_mpnet.pkl", "rb"))

print(Config.Difficulty_presentation_Full)  # type: ignore
# difficulty = int(input("Dificultad: "))
difficulty = int(input(Config.Difficulty + ": "))  # type: ignore


# with open(file_path, "r") as file:
#     secret_list = file.readlines()

# Write a function


# Optional: Remove newline characters from each element in the list
secret_list = secret_dict["basic"] if difficulty <= 2 else secret_dict["advanced"]

secret = secret_list.pop(random.randint(0, len(secret_list) - 1))
secret = secret.lower()

words = [Config.secret_word]  # type: ignore
scores = [10]

# %%
# if word not in embeddings_dict.keys():
embeddings_dict[secret] = model_st.encode(secret, convert_to_tensor=True)
model.add_vector(secret, embeddings_dict[secret].tolist())

word_vect = [embeddings_dict[secret].tolist()]

# model.add_vector(secret, embedding.tolist())

thread = threading.Thread(
    target=display_words, args=(words, pca.transform(word_vect), scores, -1)
)

# Start the thread
thread.start()


def preproc_vectors(words, word_vect, scores, repeated):
    ascending_indices = np.argsort(scores)
    # Reverse the order to get descending indices
    descending_indices = list(ascending_indices[::-1])
    ranking_data = []
    k = len(words) - 1
    if repeated != -1:
        k = repeated

    ranking_data.append(["#" + str(k), words[k], scores[k]])

    ranking_data.append("---------------------------")
    for i in descending_indices:  # type: ignore
        if i == 0:
            continue
        ranking_data.append(["#" + str(i), words[i], scores[i]])

    with open("ranking.txt", "w+") as file:
        for item in ranking_data:
            file.write("%s\n" % item)

    if len(words) > 11:
        if k in descending_indices[:11]:
            descending_indices = descending_indices[:11]
        else:
            descending_indices = descending_indices[:11]
            descending_indices.append(k)
        words_display = [words[i] for i in descending_indices]
        displayvect_display = pca.transform([word_vect[i] for i in descending_indices])
        scores_display = [scores[i] for i in descending_indices]
        bold = descending_indices.index(k)

    else:
        words_display = words
        displayvect_display = pca.transform(word_vect)
        scores_display = scores
        bold = k

    return (
        words_display,
        displayvect_display,
        scores_display,
        bold,
    )


# Example usage:

win = False
n = 0
recent_hint = 0
f_dev_avg = 0
last_hint = -1

if difficulty == 1:
    n = 3

while win == False:
    word = input(Config.New_word).lower()  # type: ignore
    if word == "give_up":
        break
    if word in words:
        repeated = words.index(word)

    else:
        repeated = -1
        words.append(word)

    thread.join()

    # if word not in embeddings_dict.keys():
    embedding = model_st.encode(word, convert_to_tensor=True)
    embeddings_dict[word] = embedding
    # model.add_vector(word, embeddings_dict[word].tolist())
    model.add_vector(word, embedding.tolist())  # type: ignore
    # model.add_vector(word, embedding.tolist())
    if repeated == -1:
        word_vect.append(embeddings_dict[word].tolist())

    score = round(model.similarity(secret, word) * 10, 2)

    if repeated == -1:
        scores.append(score)  # type: ignore
    #
    # score = round(score * 10, 2)
    # %%
    if score <= 2.5:
        feedback = Config.Feedback_0 + str(score)  # type: ignore

    elif score > 2.5 and score <= 4.0:
        feedback = Config.Feedback_1 + str(score)  # type: ignore

    elif score > 4.0 and score <= 6.0:
        feedback = Config.Feedback_2 + str(score)  # type: ignore

    elif score > 6.0 and score <= 7.5:
        feedback = Config.Feedback_3 + str(score)  # type: ignore

    elif score > 7.5 and score <= 8.0:
        feedback = Config.Feedback_4 + str(score)  # type: ignore

    elif score > 8.0 and score < 10.0:
        feedback = Config.Feedback_5 + str(score)  # type: ignore

    else:
        win = True
        feedback = Config.Feedback_8  # type: ignore
        words[0] = secret
        words.pop(len(words) - 1)
        word_vect.pop(len(word_vect) - 1)
        scores.pop(len(scores) - 1)
        # print(model.most_similar(secret, topn=20))

    print(feedback)
    if score > scores[len(scores) - 2] and win == False:
        print(Config.Feedback_6)  # type: ignore
    elif score < scores[len(scores) - 2] and win == False:
        print(Config.Feedback_7)  # type: ignore

    if difficulty != 4:
        mov_avg = calculate_moving_average(scores[1:], 5)

        # print (mov_avg)
        if len(mov_avg) > 1 and win == False:
            f_dev = calculate_tendency_slope(mov_avg)
            # print(f_dev[len(f_dev) - 3 :])
            f_dev_avg = calculate_moving_average(f_dev, 3)
            # print(f_dev_avg[len(f_dev_avg) - 3 :])
            # print(f_dev_avg)
            if f_dev_avg[len(f_dev_avg) - 1] < 0 and recent_hint == 0:
                i = random.randint(0, len(Config.hint_intro) - 1)  # type: ignore
                print("\n")
                print(Config.hint_intro[i])  # type: ignore
                hint_text, n, last_hint = hint(
                    secret,
                    n,
                    model_st,
                    last_hint,
                    lang,
                    DictWrapper(Config_full["SPA"]["Hint"])
                    if lang == 0
                    else DictWrapper(Config_full["ENG"]["Hint"]),
                )
                print(hint_text)
                recent_hint = 3

        if recent_hint != 0:
            recent_hint -= 1

    (
        words_display,
        displayvect_display,
        scores_display,
        bold_display,
    ) = preproc_vectors(words, word_vect, scores, repeated)

    if win:
        bold_display = 0

    thread = threading.Thread(
        target=display_words,
        args=(words_display, displayvect_display, scores_display, bold_display),
    )

    # Start the thread
    thread.start()

if win == False:
    print(Config.Feedback_9 + secret)  # type: ignore
    print(Config.Feedback_10)  # type: ignore

curiosity = curiosity(
    secret,
    DictWrapper(Config_full["SPA"]["Hint"])
    if lang == 0
    else DictWrapper(Config_full["ENG"]["Hint"]),
)
print(curiosity)

with open("ranking.txt", "r") as original_file:
    file_content = original_file.readlines()


new_file_name = secret + "_" + str(datetime.now())

with open("data/" + new_file_name, "w") as new_file:
    new_file.writelines(file_content[2:])

thread.join()
exit()