Sažetak
Efficient integration of data from different
sources in the digital age is crucial for the
accuracy and usability of information, and it
represents a challenge that string-matching
algorithms aim to address. As the volume of data
grows, the need for more advanced matching
techniques becomes increasingly evident. This
review categorizes algorithms into four main
groups: sequence-based, set-based, hybrid,
and phonetic algorithms. Sequence-based
algorithms, such as the Levenshtein distance and
Smith–Waterman, stand out for their precision
at the character level. Set-based methods,
including Jaccard and cosine similarity, are
effective when working with tokenized data.
Hybrid approaches combine the strengths of
these methods and offer flexibility in complex
scenarios. Phonetic algorithms, such as Soundex,
Metaphone, and NYSIIS, enable matching based
on pronunciation and are particularly useful
for names and terms with linguistic variations.
The review highlights the importance of the
continued development of all approaches to
address modern challenges of data integration.
Ključne riječi
string matching similarity measures data integration duplicate detection