Zomato Dataset  ·  4 Models  ·  v2.0

Sentiment Intelligence Platform

Decode Sentiment.

Enterprise NLP engine · RoBERTa · DistilRoBERTa · BERT · ALBERT
Fine-tuned on 2.9M+ Zomato Reviews · Three-class · Sub-second inference

Explore API Docs Health Check
4
Transformers
3
Classes
2.9M+
Reviews
512
Max Tokens
● API Online | RoBERTa Ready |
moodlens · api demo
$ curl -X POST http://localhost:8000/predict \
     -H "Content-Type: application/json" \
     -d '{"text": "Best biryani I have ever had!"}'

# 200 OK · 284ms
{
  "prediction": "Positive",
  "confidence": 0.978,
  "positive":  0.978,
  "neutral":   0.015,
  "negative":  0.007
}

$

API Endpoints

Four Routes.
Zero Confusion.

Click "Try it live" on any card — opens Swagger directly at that endpoint, ready to test with one click.

POST
/predict
Single-model RoBERTa sentiment prediction
  • Text is tokenised and truncated to 512 tokens, passed through RoBERTa in a single forward pass
  • Softmax output mapped to Positive / Neutral / Negative with raw probability for each class
  • Returns prediction label, confidence score, all three class probabilities in one clean JSON
  • Best for high-throughput pipelines where one best-in-class model is sufficient
InferenceRoBERTaJSON
Try it live
POST
/compare
All 4 models — parallel inference & consensus
  • Runs RoBERTa, DistilRoBERTa, BERT, and ALBERT on the same text independently
  • Each model's raw labels normalised to the same three-class schema before comparison
  • Results sorted by confidence — highest-confidence model ranked first in the response array
  • Use when you need cross-model validation or the most reliable possible prediction
InferenceMulti-ModelEnsemble
Try it live
GET
/health
Server status & ML engine diagnostics
  • Returns server liveness, API version, and active model list in a single JSON response
  • Lists all four active model names and training dataset for at-a-glance verification
  • Ideal for CI/CD pipelines, uptime monitors, and pre-flight checks before batch inference jobs
  • Returns UTC timestamp to verify server clock is correctly synchronised
SystemMonitoringDevOps
View live
GET
/docs
Interactive Swagger UI — test every endpoint live
  • Full OpenAPI 3.1 schema auto-generated from Pydantic models — every field typed and described
  • "Try it out" lets you fire real requests to /predict and /compare without writing any code
  • Every request and response schema documented inline with constraints and live example values
  • Dark-themed Swagger with Monokai syntax highlighting and request duration display
DocsOpenAPISwagger
Open Docs

Transformer Models

Four Engines.
One Verdict.

Each model brings a unique specialisation — together they form a robust, cross-validated ensemble.

🧠
RoBERTa
cardiffnlp/twitter-roberta-base-sentiment-latest
Default model. Trained on 124M tweets. Best general-purpose accuracy across review types and domains.
⭐ DEFAULT
DistilRoBERTa
mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis
40% faster than RoBERTa. Fine-tuned on financial news and consumer reviews. Low-latency inference.
FAST
🌍
BERT
nlptown/bert-base-multilingual-uncased-sentiment
Multilingual BERT fine-tuned in 6 languages. 5-star scale mapped to Positive / Neutral / Negative.
MULTILINGUAL
🎯
ALBERT
textattack/albert-base-v2-yelp-polarity
Parameter-efficient architecture fine-tuned on Yelp reviews. Excellent on short, punchy restaurant feedback.
EFFICIENT