oe_dictionary.py

import pandas as pd
import re
from pathlib import Path
from typing import List, Dict, Optional, Union

class OldEnglishDictionary:
    """A class to query the aligned Old English Dictionary dataset."""
    
    def __init__(self, data_path: str = "data/unified_index.tab"):
        """Initialize the dictionary with the unified index.
        
        Args:
            data_path: Path to the unified_index.tab file
        """
        self.data_path = Path(data_path)
        self.df = pd.read_csv(self.data_path, sep='\t')
        
        # Fill NaN values with empty strings for easier processing
        self.df = self.df.fillna('')
        
        # Create lowercase versions of key columns for case-insensitive search
        self.df['canonical_key_lower'] = self.df['canonical_key'].str.lower()
        self.df['normalised_form_lower'] = self.df['normalised_form'].str.lower()
        
        # Dictionary sources
        self.sources = ['BT', 'CH', 'Sweet', 'DOE', 'ParCorOE']
        
        # Count total entries
        self.total_entries = len(self.df)
        print(f"Loaded {self.total_entries} aligned entries from {self.data_path}")
    
    def search(self, query: str, 
               mode: str = 'canonical', 
               exact: bool = False, 
               pos: Optional[str] = None, 
               source: Optional[str] = None) -> pd.DataFrame:
        """Search for entries matching the query.
        
        Args:
            query: The search term
            mode: Search mode ('canonical', 'normalised', 'headword')
            exact: If True, match the entire term; if False, use substring matching
            pos: Filter by part of speech
            source: Filter by dictionary source
            
        Returns:
            DataFrame with matching entries
        """
        query = query.lower()
        
        # Base query logic
        if mode == 'canonical':
            if exact:
                results = self.df[self.df['canonical_key_lower'] == query]
            else:
                results = self.df[self.df['canonical_key_lower'].str.contains(query, regex=False)]
        
        elif mode == 'normalised':
            if exact:
                results = self.df[self.df['normalised_form_lower'] == query]
            else:
                results = self.df[self.df['normalised_form_lower'].str.contains(query, regex=False)]
        
        elif mode == 'headword':
            # Search across all dictionary headword columns
            mask = pd.Series(False, index=self.df.index)
            for source in self.sources:
                col = f"{source}_headword"
                if col in self.df.columns:
                    if exact:
                        mask |= (self.df[col].str.lower() == query)
                    else:
                        mask |= (self.df[col].str.lower().str.contains(query, regex=False, na=False))
            results = self.df[mask]
            
        else:
            raise ValueError(f"Invalid search mode: {mode}. Use 'canonical', 'normalised', or 'headword'.")
        
        # Apply filters
        if pos:
            results = results[results['POS'].str.lower() == pos.lower()]
            
        if source:
            col = f"{source}_headword"
            if col in self.df.columns:
                results = results[results[col] != '']
            else:
                raise ValueError(f"Invalid source: {source}. Use one of {self.sources}")
        
        return results
    
    def display_entry(self, entry: pd.Series) -> None:
        """Display a single dictionary entry in a readable format.
        
        Args:
            entry: A single row from the dataframe
        """
        print("\n" + "=" * 80)
        print(f"CANONICAL KEY: {entry['canonical_key']}")
        print(f"NORMALIZED FORM: {entry['normalised_form']}")
        print(f"MATCH TYPE: {entry['match_type']}")
        if entry['POS']:
            print(f"PART OF SPEECH: {entry['POS']}")
        if entry['Gloss']:
            print(f"GLOSS: {entry['Gloss']}")
        print("-" * 80)
        
        print("DICTIONARY HEADWORDS:")
        for source in self.sources:
            col = f"{source}_headword"
            if col in entry and entry[col]:
                print(f"  {source}: {entry[col]}")
        print("=" * 80)
    
    def display_results(self, results: pd.DataFrame, limit: int = 10) -> None:
        """Display search results in a readable format.
        
        Args:
            results: DataFrame containing search results
            limit: Maximum number of results to display
        """
        num_results = len(results)
        
        if num_results == 0:
            print("No matching entries found.")
            return
        
        print(f"\nFound {num_results} matching entries.")
        
        if num_results > limit:
            print(f"Displaying first {limit} results.")
            results = results.head(limit)
        
        for _, entry in results.iterrows():
            self.display_entry(entry)
    
    def find_similar(self, query: str, method: str = 'sound') -> pd.DataFrame:
        """Find entries similar to the query using various similarity methods.
        
        Args:
            query: The search term
            method: The similarity method ('sound', 'edit', 'prefix')
            
        Returns:
            DataFrame with similar entries
        """
        query = query.lower()
        
        if method == 'sound':
            # Apply Ellis vocalic transformations similar to those used in alignment
            variations = self._generate_ellis_variations(query)
            mask = pd.Series(False, index=self.df.index)
            for variation in variations:
                mask |= (self.df['canonical_key_lower'] == variation)
            return self.df[mask]
            
        elif method == 'edit':
            # Crude edit distance approximation - finding entries 
            # that share at least half their characters with the query
            query_set = set(query)
            results = []
            
            for idx, row in self.df.iterrows():
                canonical = row['canonical_key_lower']
                # Skip very short terms or terms with big length difference
                if len(canonical) < 3 or abs(len(canonical) - len(query)) > 3:
                    continue
                
                canonical_set = set(canonical)
                intersection = len(query_set.intersection(canonical_set))
                union = len(query_set.union(canonical_set))
                
                # Jaccard similarity threshold
                if intersection / union > 0.6:
                    results.append(idx)
            
            return self.df.loc[results]
            
        elif method == 'prefix':
            # Find entries that start with the query
            return self.df[self.df['canonical_key_lower'].str.startswith(query)]
            
        else:
            raise ValueError(f"Invalid method: {method}. Use 'sound', 'edit', or 'prefix'.")
    
    def _generate_ellis_variations(self, word: str) -> List[str]:
        """Generate variations based on Ellis vocalic correspondences.
        
        Args:
            word: The input word
            
        Returns:
            List of phonologically plausible variations
        """
        variations = [word]
        
        # Ellis vocalic correspondences
        transformations = [
            (r'ie', r'y'),    # fierd/fyrd
            (r'y', r'ie'),    # fyrd/fierd
            (r'ie', r'i'),    # diere/dire
            (r'i', r'ie'),    # diren/dieren
            (r'io', r'eo'),   # bion/beon
            (r'eo', r'io'),   # beorht/biorht
            (r'on', r'an'),   # monig/manig
            (r'an', r'on'),   # manig/monig
            (r'om', r'am'),   # from/fram
            (r'am', r'om'),   # fram/from
            (r'rg', r'rh'),   # burg/burh
            (r'rh', r'rg'),   # burh/burg
            (r'ea', r'a'),    # eald/ald
            (r'a', r'ea'),    # ald/eald
            (r'eo', r'e')     # eofot/efot
        ]
        
        for pattern, replacement in transformations:
            if pattern in word:
                variations.append(word.replace(pattern, replacement))
        
        return variations

    def get_statistics(self) -> Dict[str, Union[int, float]]:
        """Get statistics about the dictionary dataset.
        
        Returns:
            Dictionary with statistics
        """
        stats = {
            "total_entries": self.total_entries,
            "multi_source_entries": len(self.df[self.df['match_type'] != 'single']),
            "exact_matches": len(self.df[self.df['match_type'] == 'exact']),
            "ellis_matches": len(self.df[self.df['match_type'] == 'ellis_vocalic']),
            "single_source": len(self.df[self.df['match_type'] == 'single'])
        }
        
        # Add source-specific counts
        for source in self.sources:
            col = f"{source}_headword"
            if col in self.df.columns:
                stats[f"{source}_entries"] = len(self.df[self.df[col] != ''])
        
        # Add POS distribution
        pos_counts = self.df['POS'].value_counts().to_dict()
        stats["pos_distribution"] = pos_counts
        
        return stats


# Example usage
if __name__ == "__main__":
    oe_dict = OldEnglishDictionary()
    
    # Simple demo of the search functionality
    print("\n=== SEARCH DEMONSTRATION ===")
    
    # Search by canonical key
    print("\nSearching for entries with canonical key containing 'abelgan'...")
    results = oe_dict.search("abelgan", mode="canonical")
    oe_dict.display_results(results)
    
    # Search by normalized form
    print("\nSearching for entries with normalized form containing 'ābēodan'...")
    results = oe_dict.search("ābēodan", mode="normalised", exact=True)
    oe_dict.display_results(results)
    
    # Search by headword across all dictionaries
    print("\nSearching for entries where any dictionary has headword containing 'cyning'...")
    results = oe_dict.search("cyning", mode="headword")
    oe_dict.display_results(results, limit=5)
    
    # Filter by part of speech
    print("\nSearching for noun entries with canonical key containing 'helm'...")
    results = oe_dict.search("helm", mode="canonical", pos="N")
    oe_dict.display_results(results, limit=5)
    
    # Find similar words
    print("\nFinding entries similar to 'beon' using sound correspondences...")
    results = oe_dict.find_similar("beon", method="sound")
    oe_dict.display_results(results)
    
    # Show dictionary statistics
    stats = oe_dict.get_statistics()
    print("\n=== DICTIONARY STATISTICS ===")
    for key, value in stats.items():
        if key != "pos_distribution":
            print(f"{key}: {value}")
    
    print("\nPart of Speech Distribution:")
    for pos, count in stats["pos_distribution"].items():
        if pos:  # Skip empty POS
            print(f"  {pos}: {count}")