nlplite 0.1.2

Fast lightweight NLP library for reliable concept extraction and negation detection

NLPLite

Lightweight NLP library for fast concept extraction and enhanced negation and uncertainty detection.

Features

• Fast Aho Corasick term extraction (C-accelerated via pyahocorasick)
• NegEx Engine negation & uncertainty detection
• CLI for quick runs

---

Install

pip install nlplite

---

Quick Start

Basic Example

from nlplite import flash_extractor

# Term to code dictionary
dictionary = {"headache": "C0018681", "fever": "C0015967"}

extractor = flash_extractor(dictionary)
text = "The patient denies headache but reports fever."

print(extractor.run(text, negation_check=True))

Output:

[('headache:N', 'C0018681:N', 18, 25), ('fever:Y', 'C0015967:Y', 40, 44)]

Structured JSONL-Style Record

rec = extractor.extract_note("1234", text, negation_check=True)
print(rec)

Output:

{
  "note_id": "1234",
  "text": "The patient denies headache but reports fever.",
  "extractions": [
    ["headache:N", "C0018681:N", 18, 25],
    ["fever:Y", "C0015967:Y", 40, 44]
  ]
}

---

Dictionary Formats

A) Two-Column File (term, code)

Use when you want codes in the output (CUI/SNOMED/ICD etc)

Example CSV:

term,code
fever,C0015967
headache,C0018681

Example TSV:

term	code
fever	C0015967
headache	C0018681

Loader:

# CSV
extractor = flash_extractor("terms.csv", sep=",", header=True)

# TSV
extractor = flash_extractor("terms.tsv", sep="\t", header=True)

B) Term List (No Codes)

If you omit sep, the file is treated as a term list (one term per line).

Example:

fever
headache
shortness of breath

Loader:

extractor = flash_extractor("terms.txt")  # sep omitted → term list mode

---

Notes

• Term suffix: :N = negated, :Y = affirmed, :U = uncertain
• Uncertainty: Handled via NegEx-style phrase matching
• Case-insensitive by default
• Returns tuples: (term, code, start_pos, end_pos)