| from transformers import pipeline |
| import gradio as gr |
| import re |
|
|
| |
| ner = pipeline("ner", model="dslim/bert-base-NER", aggregation_strategy="simple") |
|
|
| |
| def split_sentences(text): |
| """ |
| Splits text into sentences while protecting abbreviations like 'U.S.' or 'Dr.'. |
| Avoids variable-length lookbehind errors by using placeholder replacement. |
| """ |
| |
| protected_terms = [ |
| "Mr.", "Mrs.", "Ms.", "Dr.", "Jr.", "Sr.", |
| "U.S.", "D.C.", "vs.", "Lt.", "St.", "Prof.", "Inc.", "Ltd.", "etc." |
| ] |
|
|
| |
| protected_map = {term: term.replace(".", "<DOT>") for term in protected_terms} |
| for original, safe in protected_map.items(): |
| text = text.replace(original, safe) |
|
|
| |
| pattern = re.compile(r"(?<=[.!?])\s+(?=[A-Z'‘“a-z])") |
| parts = re.split(pattern, text.strip()) |
|
|
| |
| restored = [] |
| for s in parts: |
| for original, safe in protected_map.items(): |
| s = s.replace(safe, original) |
| restored.append(s.strip()) |
|
|
| return [s for s in restored if s] |
|
|
| |
| def analyze_text(text): |
| sentences = split_sentences(text) |
| results = [] |
|
|
| for i, sentence in enumerate(sentences, start=1): |
| entities = ner(sentence) |
| results.append({ |
| "sentence_number": i, |
| "sentence": sentence, |
| "entities": entities |
| }) |
|
|
| return {"sentences": results, "total_sentences": len(sentences)} |
|
|
| |
| demo = gr.Interface( |
| fn=analyze_text, |
| inputs=gr.Textbox(label="Input Text", lines=6, placeholder="Paste your article..."), |
| outputs=gr.JSON(label="NER + Sentence Output"), |
| title="Sentence Splitter + NER API", |
| description="Splits text into sentences (protects abbreviations) and runs Named Entity Recognition (dslim/bert-base-NER)." |
| ) |
|
|
| if __name__ == "__main__": |
| demo.launch() |
|
|
