149 lines
5.2 KiB
Python
149 lines
5.2 KiB
Python
import math
|
|
import re
|
|
from collections import Counter
|
|
|
|
|
|
class UKWMSAI_RAG:
|
|
"""Sistem RAG Sederhana berbasis Pure Python untuk topik Kecerdasan Buatan
|
|
|
|
di Teknik Informatika Universitas Katolik Widya Mandala Surabaya (UKWMS).
|
|
"""
|
|
|
|
def __init__(self, knowledge_base):
|
|
print("[INFO] Menginisialisasi Basis Pengetahuan UKWMS - AI...")
|
|
self.documents = knowledge_base
|
|
self.vocab = self._build_vocab(self.documents)
|
|
# Pra-penghitungan vektor TF-IDF untuk dokumen lokal
|
|
self.doc_vectors = [
|
|
self._compute_tfidf(doc, self.vocab, self.documents)
|
|
for doc in self.documents
|
|
]
|
|
|
|
def _tokenize(self, text):
|
|
"""Membersihkan teks dan memecahnya menjadi kata-kata (token)."""
|
|
return re.findall(r"\b\w+\b", text.lower())
|
|
|
|
def _build_vocab(self, docs):
|
|
"""Membangun kamus kata unik dari seluruh dokumen."""
|
|
vocab = set()
|
|
for doc in docs:
|
|
vocab.update(self._tokenize(doc))
|
|
return sorted(list(vocab))
|
|
|
|
def _compute_tfidf(self, text, vocab, all_docs):
|
|
"""Menghitung bobot TF-IDF secara manual tanpa library eksternal."""
|
|
tokens = self._tokenize(text)
|
|
tf = Counter(tokens)
|
|
total_tokens = len(tokens) if tokens else 1
|
|
|
|
vector = {}
|
|
for word in vocab:
|
|
# Term Frequency (TF)
|
|
term_freq = tf[word] / total_tokens
|
|
|
|
# Inverse Document Frequency (IDF)
|
|
doc_count = sum(1 for d in all_docs if word in self._tokenize(d))
|
|
idf = (
|
|
math.log(len(all_docs) / (doc_count + 1)) + 1 if doc_count > 0 else 0
|
|
)
|
|
|
|
vector[word] = term_freq * idf
|
|
return vector
|
|
|
|
def _cosine_similarity(self, vec1, vec2):
|
|
"""Menghitung kemiripan kosinus antara dua vektor TF-IDF."""
|
|
intersection = set(vec1.keys()).intersection(set(vec2.keys()))
|
|
numerator = sum(vec1[x] * vec2[x] for x in intersection)
|
|
|
|
sum1 = sum(val ** 2 for val in vec1.values())
|
|
sum2 = sum(val ** 2 for val in vec2.values())
|
|
denominator = math.sqrt(sum1) * math.sqrt(sum2)
|
|
|
|
if denominator == 0:
|
|
return 0.0
|
|
return numerator / denominator
|
|
|
|
def retrieve(self, query, top_k=1):
|
|
"""Mencari dokumen paling relevan menggunakan Cosine Similarity."""
|
|
query_vector = self._compute_tfidf(query, self.vocab, self.documents)
|
|
scores = []
|
|
|
|
for i, doc_vec in enumerate(self.doc_vectors):
|
|
score = self._cosine_similarity(query_vector, doc_vec)
|
|
scores.append((score, self.documents[i]))
|
|
|
|
# Urutkan berdasarkan skor tertinggi
|
|
scores.sort(key=lambda x: x[0], reverse=True)
|
|
return [doc for score, doc in scores[:top_k]]
|
|
|
|
def generate_answer(self, query):
|
|
"""Tahap Generation: Menggabungkan hasil retrieval dengan format jawaban RAG."""
|
|
retrieved_docs = self.retrieve(query, top_k=1)
|
|
context = (
|
|
retrieved_docs[0]
|
|
if retrieved_docs
|
|
else "Maaf, informasi tidak ditemukan."
|
|
)
|
|
|
|
# Template Generator Extractive RAG
|
|
answer = (
|
|
f"==================================================\n"
|
|
f"HASIL GENERASI RAG - INFORMATIKA UKWMS (KECERDASAN BUATAN)\n"
|
|
f"==================================================\n"
|
|
f"Pertanyaan : {query}\n\n"
|
|
f"Konteks Relevan (Retrieved):\n-> {context}\n\n"
|
|
f"Jawaban Sistem:\n"
|
|
f"Berdasarkan kurikulum dan data akademik Prodi Informatika UKWMS, "
|
|
f"fokus pada bidang Kecerdasan Buatan mencakup informasi berikut:\n"
|
|
f'"{context}"\n'
|
|
f"=================================================="
|
|
)
|
|
return answer, retrieved_docs
|
|
|
|
|
|
# ==========================================
|
|
# EKSEKUSI PROGRAM
|
|
# ==========================================
|
|
if __name__ == "__main__":
|
|
# 1. Dataset / Basis Pengetahuan Lokal UKWMS - Kecerdasan Buatan
|
|
ukwms_ai_kb = [
|
|
(
|
|
"Program Studi Informatika Fakultas Teknik UKWMS menyediakan"
|
|
" peminatan khusus Data Science and Artificial Intelligence (AI)"
|
|
" untuk membekali mahasiswa dengan kemampuan analisis data dan"
|
|
" pembelajaran mesin."
|
|
),
|
|
(
|
|
"Mata kuliah inti dalam bidang Kecerdasan Buatan di Informatika UKWMS"
|
|
" meliputi Machine Learning, Deep Learning, Natural Language"
|
|
" Processing (NLP), Computer Vision, dan Data Mining."
|
|
),
|
|
(
|
|
"Laboratorium Komputasi dan Sistem Cerdas di Teknik Informatika"
|
|
" UKWMS Kampus Kalijudan Surabaya sering digunakan mahasiswa untuk"
|
|
" mengembangkan proyek akhir berbasis kecerdasan buatan dan IoT."
|
|
),
|
|
(
|
|
"Penelitian dosen dan mahasiswa Informatika UKWMS di bidang AI"
|
|
" banyak berfokus pada pengenalan pola citra medis, sistem"
|
|
" rekomendasi, dan analisis sentimen bahasa Indonesia."
|
|
),
|
|
(
|
|
"Lulusan peminatan Kecerdasan Buatan dari Informatika UKWMS bergelar"
|
|
" Sarjana Komputer (S.Kom.) dengan profisiensi tinggi sebagai AI"
|
|
" Engineer, Data Scientist, dan Machine Learning Practitioner."
|
|
),
|
|
]
|
|
|
|
# 2. Inisialisasi Sistem RAG
|
|
rag = UKWMSAI_RAG(ukwms_ai_kb)
|
|
|
|
# 3. Contoh Pengujian Pertanyaan Tugas
|
|
sample_query = (
|
|
"Apa saja mata kuliah kecerdasan buatan yang dipelajari di informatika"
|
|
" ukwms?"
|
|
)
|
|
|
|
# 4. Menjalankan Pipeline
|
|
final_answer, sources = rag.generate_answer(sample_query)
|
|
print("\n" + final_answer) |