Den Vektorraum untersuchen

Einführung in Embeddings mit der OpenAI API

Emmanuel Pire

Senior Software Engineer, DataCamp

Beispiel: Schlagzeilen einbetten

articles = [
    {"headline": "Economic Growth Continues Amid Global Uncertainty", "topic": "Business"},
    {"headline": "Interest rates fall to historic lows", "topic": "Business"},
    {"headline": "Scientists Make Breakthrough Discovery in Renewable Energy", "topic": "Science"},
    {"headline": "India Successfully Lands Near Moon's South Pole", "topic": "Science"},
    {"headline": "New Particle Discovered at CERN", "topic": "Science"},
    {"headline": "Tech Company Launches Innovative Product to Improve Online Accessibility", "topic": "Tech"},
    {"headline": "Tech Giant Buys 49% Stake In AI Startup", "topic": "Tech"},
    {"headline": "New Social Media Platform Has Everyone Talking!", "topic": "Tech"},
    {"headline": "The Blues get promoted on the final day of the season!", "topic": "Sport"},
    {"headline": "1.5 Billion Tune-in to the World Cup Final", "topic": "Sport"}
]

Beispiel: Schlagzeilen einbetten

Eine Liste von Dictionaries mit den Schlagzeilen und ihren Embeddings.

Mehrere Eingaben einbetten

headline_text = [article['headline'] for article in articles]
headline_text

["Economic Growth Continues Amid Global Uncertainty",
 ...,
 "1.5 Billion Tune-in to the World Cup Final"]

response = client.embeddings.create(
  model="text-embedding-3-small",
  input=headline_text
)
response_dict = response.model_dump()

Batching ist effizienter als mehrere API-Calls zu machen

[...]

'data': [
    {
      "embedding": [-0.017142612487077713, ..., -0.0012911480152979493],
      "index": 0,
      "object": "embedding"
    },
    {
      "embedding": [-0.032995883375406265, ..., -0.0028605300467461348],
      "index": 1,
      "object": "embedding"
    },
    ...
  ]

[...]

Mehrere Eingaben einbetten

articles = [
    {"headline": "Economic Growth Continues Amid Global Uncertainty", "topic": "Business"},
     ...
]

for i, article in enumerate(articles):

    article['embedding'] = response_dict['data'][i]['embedding']


print(articles[:2])

[{'headline': 'Economic Growth Continues Amid Global Uncertainty',
  'topic': 'Business',
  'embedding': [-0.017142612487077713, ..., -0.0012911480152979493]}
 {'headline': 'Interest rates fall to historic lows',
  'topic': 'Business',
  'embedding': [-0.032995883375406265, ..., -0.0028605300467461348]}]

Wie lang ist der Embeddings-Vektor?

„Economic Growth Continues Amid Global Uncertainty"

len(articles[0]['embedding'])

„Tech Company Launches Innovative Product to Improve Accessibility"

len(articles[5]['embedding'])

Gibt immer 1536 Zahlen zurück!

Dimensionsreduktion und t-SNE

Verschiedene Techniken zur Reduktion der Dimensionen
t-SNE (t-distributed Stochastic Neighbor Embedding)

¹ https://www.datacamp.com/tutorial/introduction-t-sne

t-SNE implementieren

from sklearn.manifold import TSNE
import numpy as np


embeddings = [article['embedding'] for article in articles]


tsne = TSNE(n_components=2, perplexity=5)

embeddings_2d = tsne.fit_transform(np.array(embeddings))

n_components: Anzahl der Ziel-Dimensionen
perplexity: vom Algorithmus genutzt, muss kleiner als die Anzahl Datenpunkte sein
Führt zu Informationsverlust

¹ https://www.datacamp.com/tutorial/introduction-t-sne

Embeddings visualisieren

import matplotlib.pyplot as plt

plt.scatter(embeddings_2d[:, 0], embeddings_2d[:, 1])


topics = [article['topic'] for article in articles]
for i, topic in enumerate(topics):
    plt.annotate(topic, (embeddings_2d[i, 0], embeddings_2d[i, 1]))

plt.show()

Embeddings visualisieren

Ähnliche Artikel liegen beieinander!
Das Modell hat die semantische Bedeutung erfasst

Als Nächstes: Ähnlichkeit berechnen

Ein Plot des 2D-Vektorraums, der zeigt, dass Rezensionen mit gleichem Sentiment und Thema im Vektorraum näher beieinander liegen.

Lass uns üben!

Einführung in Embeddings mit der OpenAI API