Analysis and safety engineering of fuzzy string matching algorithms

In this paper we explore fuzzy string matching in an automatic ticket classification and processing system. We compare performance of the following string similarity algorithms: Longest Common Subsequence (LCS), Dice coefficient, Cosine Similarity, Levenshtein (edit) distance and Damerau distance. T...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:ISA transactions Jg. 113; S. 1 - 8
Hauptverfasser: Pikies, Malgorzata, Ali, Junade
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Elsevier Ltd 01.07.2021
Schlagworte:
Binary classification
Fuzzy string matching
Natural language processing
Neural network
Safety engineering
String similarity
String similarity
Fuzzy string matching
Natural language processing
Neural network
Safety engineering
Binary classification
ISSN:0019-0578, 1879-2022, 1879-2022
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In this paper we explore fuzzy string matching in an automatic ticket classification and processing system. We compare performance of the following string similarity algorithms: Longest Common Subsequence (LCS), Dice coefficient, Cosine Similarity, Levenshtein (edit) distance and Damerau distance. Through optimisation, we accomplished a 15% improvement in the ratio of false positives to true positive classifications over the existing approach used by a customer support system for free customers. To introduce greater safety; we compliment fuzzy string matching algorithms with a second layer Convolutional Neural Network (CNN) binary classifier, achieving an improved keyword classification ratio for two ticket categories by a relative 69% and 78%. Such an approach allows for classification to only be applied where a desired level of safety achieved, such as in instances where automated answers.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0019-0578
1879-2022
1879-2022
DOI:10.1016/j.isatra.2020.10.014