Antwort-Extraktion

Einführung in Generative KI in Snowflake

James Cha-Earley

Senior Developer Advocate, Snowflake

Warum Antwort-Extraktion?

Beschwerden erkennen
Feedback zu Ausstattung extrahieren
Fragen aus langen Dokumenten beantworten

Ein Dokument mit einem Fragezeichen, umgeben von Emojis mit lachenden oder traurigen Gesichtern – wir wissen nicht, worum es in den einzelnen Bewertungen geht

Ein häufiges Geschäftsproblem

review = """
The property needs updating and management needs to hire more staff. 
The staff does everything they can but it is not enough for the number of guests. 
The rooms look old as well as everything in the bathrooms. 
Definitely better options in the hotel zone.
"""

Snowflake Cortex hilft!

Eine Antwort extrahieren

from snowflake.cortex import extract_answer


response = extract_answer(
    from_text=review,
    question="What did the guest complain about?"
)

Ausgabe

print(response)

[
    {
    "answer": "Management needs to hire more staff.",
    "score": 0.8013839254
    }
]

Scores von 0 bis 1

Einen Workflow aufbauen

SELECT DESCRIPTION
FROM HOTELS.REVIEWS
WHERE RATING < 6
    AND DATE = '2023-08-23';

SQL-Ausgabe

I will never book a vacation through Expedia not only did I pay over 5;000 dollars
but they false advertise the resort; in my reservation It clearly stated that
breakfast and dinner was included. When I got to the resort they informed me that 
these issues have happed before with Expedia because the resort does not offer an
all exclusive. After flighting practically all day I called Expedia and they had me
on hold for 45min and nothing got resolved. I ended up spending over 1000.00 more
just on food. I am beyond me they are not only false advertising but refused to
take any responsibility on their part for the issues they caused me and my family.  Again I will never chose Expedia to plan a vacation!!!

Eine Frage stellen

negative_review = dataframe_1.to_pandas()['DESCRIPTION'].iloc[0]


response = extract_answer(
    from_text=negative_review,
    question="What was the primary issue that the guest mentioned?"
)

print(response)

Antwort prüfen

[
    {
    "answer" : "Expedia falsely advertised that meals were included, but the
                resort did not offer an all-inclusive package.",
    "score": 0.75829312442
    }
]

Einschränkungen

Liefert teils unvollständige oder vage Phrasen
Am besten für Absätze, nicht für mehrthematiche Dokumente
Kritische Ausgaben immer prüfen

Warnschild

Lass uns üben!

Einführung in Generative KI in Snowflake