kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Comparing Methods of Open Intent Discovery i n Customer Supp ort Queries
KTH, School of Electrical Engineering and Computer Science (EECS).
KTH, School of Electrical Engineering and Computer Science (EECS).
2022 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
En komparativ studie av metoder för avsiktsidentifiering i kundsupportsförfrågningar (Swedish)
Abstract [en]

Advances in technology are driving digital transformation of customer support functions. In a digital setting, consumers interact via web pages, search fields, and support channels, and user objectives are thus expressed as queries. In this paper, we investigate methods related to discovering the user’s intention behind these queries, with a focus on failed customer support queries, i.e., queries that have not been responded to by the company. Although most previous work on indent identification has been conducted on classification of queries, we intend to identify intent without prior knowledge of intent classes, known as open intent discovery. We propose a probabilistic approach, based on term frequency and bigram probability. To evaluate the performance of the method, it is compared to a dependency parser approach to open intent discovery. Both methods are applied to a dataset containing failed customer support queries and the benchmark dataset SNIPS. We have shown that the term frequency method outperforms the dependency parser approach on all datasets. Additionally, this paper identifies factors as to why automation of support functions might be needed, and the key challenges, both technical and social, as to why they might fail.

Abstract [sv]

Den teknologiska utvecklingen driver förändring där kundsupport och kommunikationskanaler blir allt mer automatiserade. I denna digitala miljö interagerar konsumenter med företag genom textförfrågningar. I denna studie undersöker vi metoder för identifiering av avsikten i sådana textförfrågningar. Traditionellt sätt har detta undersökts som ett klassificeringsproblem, men i denna studie undersöker vi användarnas avsikter utanför redan kända avsiktskategorier, så kallad öppen avsiktsidentifiering. Vi föreslår en metod baserad på sannolikhet och ordfrekvens och jämför denna med annan metod som extraherar substantiv och verb som avsiktisidentifierare. Metoderna utvärderas genom applicering på två dataset, ett innehåller ej besvarade förfrågningar riktade till kundsupport och det andra utgörs av benchmark-datasetet SNIPS. Vi har visat att ordfrekvensmetoden presterar bättre på alla dataset. Utifrån ett företagsperspektiv identifierar vi också faktorer som påverkar automatiseringsprocessen, samt vilka utmaningar som följer av detta.

Place, publisher, year, edition, pages
2022. , p. 11
Series
TRITA-EECS-EX ; 2022:315
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-319895OAI: oai:DiVA.org:kth-319895DiVA, id: diva2:1702349
Supervisors
Examiners
Available from: 2022-10-11 Created: 2022-10-10 Last updated: 2022-10-11Bibliographically approved

Open Access in DiVA

No full text in DiVA

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar

urn-nbn

Altmetric score

urn-nbn
Total: 97 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf