---
tags:
- text2text-generation
- financial-nlp
- time-extraction
- t5
language: en
license: apache-2.0
widget:
- text: "extract time: What was Apple's revenue for Q3 2023?"
  example_title: "Quarterly Period"
- text: "extract time: Show me Tesla's gross margin for fiscal year 2022"
  example_title: "Annual Period"
- text: "extract time: Get Microsoft's earnings for the third quarter of 2021"
  example_title: "Quarter with Year"
---

# T5 Time Period Extractor

This model extracts time periods from financial queries.

## Usage

```python
from transformers import pipeline

pipe = pipeline("text2text-generation", model="ak837/t5-time-period-extractor")

# Important: Always prefix with "extract time: "
result = pipe("extract time: What was Apple's revenue for Q3 2023?")
print(result[0]['generated_text'])
# Output should be JSON like: {"period":"quarter","year":2023,"quarter":3}
```

## Training

This model was fine-tuned to extract time periods from financial queries in a structured JSON format.

## Output Format

The model outputs JSON with:
- `period`: "quarter" or "annual"
- `year`: The year (e.g., 2023)
- `quarter`: Quarter number (1-4) if applicable

## Known Issue

The model sometimes outputs JSON without proper braces. You may need to add them:
```python
output = result[0]['generated_text']
if not output.startswith('{'):
    output = '{' + output + '}'
```