Powered by OpenAIRE graph
Found an issue? Give us feedback
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/ Publikationer från K...arrow_drop_down
image/svg+xml art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos Open Access logo, converted into svg, designed by PLoS. This version with transparent background. http://commons.wikimedia.org/wiki/File:Open_Access_logo_PLoS_white.svg art designer at PLoS, modified by Wikipedia users Nina, Beao, JakobVoss, and AnonMoos http://www.plos.org/
Publikationer från KTH
Bachelor thesis . 2023
addClaim

This Research product is the result of merged Research products in OpenAIRE.

You have already added 0 works in your ORCID record related to the merged Research product.

Metod för ett automatiserat frågebesvarande i det svenska språket

Authors: Penna, Kristian;

Metod för ett automatiserat frågebesvarande i det svenska språket

Abstract

I ärendehanteringssystem utgör avslutade ärenden en värdefull datamängd bestående av par av frågor och svar som organisationer med rätt metoder kan dra nytta av för att utvinna fördelar. I denna studie har en Sentence Transformers-modell blivit finjusterad för question answering som tillsammans med en datamängd från ett ärendehanteringssystem automatiskt kan besvara organisationsspecifika frågor i det svenska språket. Modellen bygger på en semantisk jämförelsemekanik för att identifiera vilken tidigare fråga i datamängden som är mest lik en ny fråga, för att således generera den tidigare frågans korresponderande svar som utdata. Den datamängden som användes till modellen bestod av 75 par av frågor och svar från ett ärendehanteringssystem. Inför testningen av modellen skapades 61 testfrågor där varje fråga var helt semantiskt lika minst en fråga som förekom bland de 75 frågorna i datamängden, samtidigt som de skiljde sig i övrigt gällande meningsuppbyggnader och inkluderade termer. Vid testningen användes testfrågorna för att mäta modellens förmåga att besvara frågor korrekt samt för att mäta prestandan på jämförelsemekaniken. Testresultaten visade att modellen kunde besvara frågor med en pricksäkerhet på 78,7%, och att jämförelsemekaniken kunde med en pricksäkerhet på 67,2% para ihop semantiskt helt lika frågor. Closed tickets within ticket management systems form a valuable dataset consisting of pairs of questions and answers that organizations can use in different ways and benefit from. In this study a Sentence Transformers model has been fine-tuned to use such a dataset for question answering to be able to automatically answer organization-specific questions in Swedish. The model is based on a semantic comparison mechanism to identify which previous question in the dataset is most similar to a new question, in order to generate the corresponding answer to the previous question as output. The dataset used for the model consisted of 75 pairs of questions and answers from a ticket management system. Before testing the model, 61 test questions were created where each question was semantically identical to at least one question that appeared among the 75 questions in the dataset, while at the same time differing in terms of sentence structure and included terms. During testing, the test questions were used to measure the model's ability to correctly answer questions and the performance of the comparison mechanism. The results showed that the model could answer questions with an accuracy of 78.7 %, and that the comparison mechanism could match semantically identical questions with an accuracy of 67.2 %.

Country
Sweden
Related Organizations
Keywords

semantic textual similarity, artificiell intelligens, Computer Sciences, sentence transformers, deep learning, artificial intelligence, språkteknologi, Language Technology (Computational Linguistics), machine learning, Datavetenskap (datalogi), maskininlärning, question answering, frågebesvarande, transformer, semantisk textlikhet, natural language processing, artificial neural networks, Språkteknologi (språkvetenskaplig databehandling), djupinlärning, BERT, artificiella neurala nätverk

4.1.3 [14] Reimers N, Gurevych I. Sentence-BERT: Sentence embeddings using Siamese BERT-networks.

arXiv [csCL] [Internet]. 2019 [citerad 2022-12-19]; Hämtad från: http://arxiv.org/abs/1908.10084 [15] SentenceTransformers documentation - sentence-transformers documentation [Internet].

Sbert.net. [citerad 2022-12-15]. Hämtad från: https://www.sbert.net/ [16] Basavarajaiah M. Maxpooling vs minpooling vs average pooling [Internet]. Medium. 2019 [citerad 2022-12-12]. Hämtad från: https://medium.com/@bdhuma/which-pooling-method-is-bettermaxpooling-vs-minpooling-vs-average-pooling-95fb03f45a9 [17] Espejel EO. Train and Fine-Tune Sentence Transformers Models [Internet]. Huggingface.co.

[citerad 2022-12-13]. Hämtad från: https://huggingface.co/blog/how-to-train-sentence-transformers [18] Majumder G, Pakray P, Gelbukh A, Pinto D. Semantic textual similarity methods, tools, and applications: A survey. Comput Sist [Internet]. 2016 [citerad 2022-12-15];20(4):647-65. Hämtad från: https://www.scielo.org.mx/scielo.php?pid=S1405-55462016000400647&script=sci_arttext [19] Nurmanbetov D. Cutting edge semantic search and sentence similarity [Internet]. Towards Data Science. 2020 [citerad 2023-01-05]. Hämtad från: https://towardsdatascience.com/cuttingedge-semantic-search-and-sentence-similarity-53380328c655 [20] Skärvad P, Olsson J. Företagsekonomi 100 Faktabok. 18:e upplagan. Stockholm: Liber; 2017.

[21] Gulliksson H, Holmgren U. Hållbar utveckling: teknik, samhälle och livskvalitet. Tredje upplagan. Lund: Studentlitteratur; 2018.

[22] Gröndahl F, Svanström M. Hållbar utveckling: en introduktion för ingenjörer och andra problemlösare. 1. uppl. Stockholm: Liber; 2011.

  • BIP!
    Impact byBIP!
    citations
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
  • citations
    This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    0
    popularity
    This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
    Average
    influence
    This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
    Average
    impulse
    This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
    Average
    Powered byBIP!BIP!
Powered by OpenAIRE graph
Found an issue? Give us feedback
citations
This is an alternative to the "Influence" indicator, which also reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Citations provided by BIP!
popularity
This indicator reflects the "current" impact/attention (the "hype") of an article in the research community at large, based on the underlying citation network.
BIP!Popularity provided by BIP!
influence
This indicator reflects the overall/total impact of an article in the research community at large, based on the underlying citation network (diachronically).
BIP!Influence provided by BIP!
impulse
This indicator reflects the initial momentum of an article directly after its publication, based on the underlying citation network.
BIP!Impulse provided by BIP!
0
Average
Average
Average