turkish-text-classification-model

https://huggingface.co/algumusrende/turkish-text-classification-model

This model is used for Sentiment Analysis, which is based on bert-base-turkish-sentiment-cased https://huggingface.co/savasy/bert-base-turkish-sentiment-cased

Dataset

The dataset is taken from https://www.kaggle.com/datasets/burhanbilenn/duygu-analizi-icin-urun-yorumlari?select=magaza_yorumlari_duygu_analizi.csv

Containing product reviews of electronics stores in Turkish Language, with 3 categories:

[ "Olumlu (Positive)", "Olumsuz (Negative)", "Tarafsız (Neutral)" ]

2 columns and 11429 rows (3 NaN rows), encoded in "utf-16"

Dataset

size	data
5713	train.csv
2856	val.csv
2857	test.tsv
11426	total

Training and Results

index	eval_loss	eval_Accuracy	eval_F1	eval_Precision	eval_Recall
train	0.41672539710998535	0.8531419569403116	0.8346503162224169	0.842628684710363	0.8315839726920476
val	0.6787932515144348	0.7545518207282913	0.7277930570101517	0.7311753495947505	0.7293434379700242
test	0.6885481476783752	0.7434371718585929	0.7170880702233838	0.7189901255561661	0.7180628887201386

Code Usage

from transformers import AutoModelForSequenceClassification, AutoTokenizer, pipeline

model = AutoModelForSequenceClassification.from_pretrained("algumusrende/turkish-text-classification-model")
tokenizer= AutoTokenizer.from_pretrained("algumusrende/turkish-text-classification-model")
pipe = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)

pipe("Son zamanlarda ekonomideki istikrar, borsa endeksine de olumlu yansıdı.")
# [{'label': 'Olumlu', 'score': 0.6654265522956848}]

pipe("Geçirdiğim diş operasyonu için çekilen röntgen filmleri sağlık yardımı kapsamında ödenmedi.")
# [{'label': 'Olumsuz', 'score': 0.9064584970474243}]

pipe("Eskiden bayramlarda çikolata dağıtlırdı, artık bunu göremiyoruz.")
# [{'label': 'Olumsuz', 'score': 0.7049197554588318}]

pipe("Ürün genel itibari ile iyi sayılır, ancak bazı eksikleri de var.")
# [{'label': 'Tarafsız', 'score': 0.9369649887084961}]

---