Ähnliche strings finden (wie z.B. google Suche)

[himitsu]

Code:

text = ["Meier", "Maier", "Mayer"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))

text = ["tough", "though"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))

text = ["Baum", "Tanne"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))

text = ["Nadelbaum", "Laubbaum"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))

text = ["Hamster", "Goldfisch", "Schäferhund", "Dackel"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))
print(f"{text[2]} {text[3]} ", cosine_similarity(test[2], test[3]))

text = ["Mann", "Frau", "Junge", "Bube", "Auto", "Haus"]
test = embedding_model.encode(text)
print(f"{text[0]} {text[1]} ", cosine_similarity(test[0], test[1]))
print(f"{text[2]} {text[3]} ", cosine_similarity(test[2], test[3]))
print(f"{text[4]} {text[5]} ", cosine_similarity(test[4], test[5]))

Code:

intfloat/multilingual-e5-large

Meier Maier  0.93333364
tough though  0.86950207
Baum Tanne  0.79528135
Nadelbaum Laubbaum  0.8690178
Hamster Goldfisch  0.80143225
Schäferhund Dackel  0.786349
Mann Frau  0.86286455
Junge Bube  0.83770347
Auto Haus  0.8216029

Ehrlich gesagt, hätte ich die Hunde in der Ähnlichkeit höher erwartet, als die anderen Tiere.

OK, passt doch

Code:

Alignment-Lab-AI/e5-mistral-7b-instruct

Meier Maier  0.93829346
tough though  0.7688065
Baum Tanne  0.90267164
Nadelbaum Laubbaum  0.88353676
Hamster Goldfisch  0.78831625
Schäferhund Dackel  0.844885
Mann Frau  0.84775054
Junge Bube  0.8940252
Auto Haus  0.67513704

Achtung, Alignment-Lab-AI/e5-mistral-7b-instruct benötigt 32 GB RAM.