Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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
Comparison between statistical and neuronal models for machine translation
KTH, School of Electrical Engineering and Computer Science (EECS).
2018 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesis
Abstract [en]

Machine translation is a thriving field that deals with multiple of the challenges that the modern world face. From accessing to knowledge in a foreign language, to being able to communicate with people that does not speak the language, we can take great benefit from automatic translation made by software. The state-of-the-art models of machine translation during the last decades, based of inferred statistical knowledge over a set of parallel data, had been recently challenged by neural models based on large artificial neural networks.

This study aims to compare both methods of machine translation, the one based on statistical inference (SMT) and the one based on neural networks (NMT). The objective of the project is to compare the performance and the computational needs of both models depending on different factors like the size of the training data or the likeliness of language pair.

To make this comparison I have used publicly available parallel data and frameworks in order to implement the models. The evaluations of said models are done under the BLEU score, which computes the correspondence of the translation with the translation made by a human operation.

The results indicate that the SMT model outperforms the NMT model given relatively small amount of data and a basic set of techniques. The results also shown that NMT have a substantially higher need of processing power, given that the training of large ANN is more demanding than the statistical inference.

Place, publisher, year, edition, pages
2018.
Series
TRITA-EECS-EX ; 2018:224
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-239557OAI: oai:DiVA.org:kth-239557DiVA, id: diva2:1265824
Supervisors
Examiners
Available from: 2018-11-26 Created: 2018-11-26 Last updated: 2018-11-26Bibliographically 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: 3 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • 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