Added NLP-Based Section Matching and Data Extraction Logic Proof of Concept

[SiriChandanaGarimella]

Fixes #99

What was changed?

1. The section matching logic in the ORCALogProcessor class is implemented.
2. Enhanced NLP capabilities were added using:

• Text normalization (lowercasing and lemmatization with spaCy).
• Fuzzy matching for near-similar terms (e.g., "gibbs free energies" → "gibbs free energy").

3. Updated _nlp_match_section and added a helper function _lemmatize_term

Why was it changed?
The current rule-based system needs to be replaced with a more robust solution that can maintain accuracy, scale easily and efficiently, and reduce maintenance. Based on the analysis from Issue #98, implemented an NLP-based system for extracting data from ORCA log files as a proof of concept.

How was it changed?
Files added: orca_log_processor.py - Contains the implementation of the ORCALogProcessor class.

Key Functionalities Added:
Using Natural Language Processing (NLP), improved data extraction from text files, particularly when dealing with variations in section names (e.g., "INNER ENERGIES" vs "INNER ENERGY"). Here’s what is done:

1. Normalization of Text:
   • Text is converted to lowercase to ensure case insensitivity during matching.
   • Words are lemmatized using spaCy, which reduces words to their base form (e.g., "energies" → "energy").
2. Fuzzy Matching:
   • The similarity between terms is calculated using SequenceMatcher from Python's difflib module. This helps match near-similar words with a threshold of 90% (e.g., "gibbs free energies" vs "gibbs free energy").
3. Multi-Word Matching:
   • Multi-word terms like "gibbs free energies" are split and lemmatized word-by-word for robust comparison.
4. Section Matching:
   • The extracted and normalized search term is compared against known section patterns (self.section_patterns) to find the correct section.
5. Cycle Boundaries Detection:
   • Automatically detects start and end lines of GEOMETRY OPTIMIZATION CYCLES using regex.

Future Improvements for Data Extraction
As this is a Proof of Concept (POC), further improvements can include:

1. Advanced NLP Models: Use more advanced NLP models like BERT or GPT to identify and extract sections based on semantic meaning, not just exact/fuzzy matches.
2. Keyword Extraction: Automatically extract keywords and patterns from text files instead of hardcoding section names.
3. Regular Expression Enhancements: Use regex to detect section headers more efficiently.
4. User Input Flexibility: Allow users to specify custom similarity thresholds or patterns dynamically.
5. Error Handling and Logs: Provide detailed logs for unmatched terms and offer suggestions for valid inputs.
6. Visualization: Visualize matched sections or highlight matched keywords for better readability.
7. Integration with Machine Learning: Train a classifier to automatically detect and group sections from new datasets.
8. Enhance performance for very large log files.

Screenshot
Input:
search_term: GIBBS FREE ENERGIES
lines_specified: FIRST 30
sections: 1, 2, 3

Preview Document Output: