Metaheuristics for Tuning Model Parameters in Two Natural Language Processing Applications

Kłyk, Łukasz; Myszkowski, Paweł B.; Broda, Bartosz; Piasecki, Maciej; Urbansky, David

doi:10.1007/978-3-642-33185-5_4

Cited by 2 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The value of τ 0 depends on the KSs, but it can be set to the smallest weight value that signals good triples in the KS of the biggest coverage. All threshold values can be also automatically optimised, e.g., as in [43].…”

Section: Algorithmmentioning

confidence: 99%

“…It can block some pre-defined junctions of links of selected types or at least decrease the amount of activation going through link junctions of certain types, e.g. holonymy, antonymy The value of was heuristically set to τ 0 /4, but it can be obtained during optimisation, as all other parameters, cf [43]. The parameters µ and together control the maximal distance on which the initial activation of a node can influence its local subgraph.…”

Section: Activation Replicationmentioning

confidence: 99%

“…It shows the versatility of the model. The values of the transmittance coefficient can be optimised in a way similar to the optimisation of the other parameters, cf [43].…”

Section: Transmittance and Impedance Implementationmentioning

confidence: 99%

See 2 more Smart Citations

Paintball – Automated Wordnet Expansion Algorithm based on Distributional Semantics and Information Spreading

Piasecki¹

2019

CMST

Self Cite

View full text Add to dashboard Cite

plWordNet has been consequently built on the basis of the corpus-based wordnet development method. As plWordNet construction had started from scratch it was necessary to find a way to reduce the amount of work required, and not to reduce the quality. In the paper we discuss the gained experience in applying different tools based on Distributional Semantics methods to support the work of lexicographers. A special attention is given to the Paintball algorithm for semiautomated wordnet expansion and its application in the WordnetWeaver system.

show abstract

Section: Algorithmmentioning

confidence: 99%

Section: Activation Replicationmentioning

confidence: 99%

See 1 more Smart Citation

Paintball – Automated Wordnet Expansion Algorithm based on Distributional Semantics and Information Spreading

Piasecki¹

2019

CMST

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the state-of-the-art results for Polish suggest that further improvements of measures of semantic relatedness are still possible, for example by using a constraint-based approach, a dependency parser, and testing more measures with more parameters. Similarly, the attachment algorithm could further be improved by optimizing parameters of the algorithms, for example by using meta-heuristics like in (Kłyk et al 2012), Slovenščina 2.0, 2 (2013 [108]…”

Section: O N C L U S I O N Smentioning

confidence: 99%

Utemeljevanje sloWNeta na korpusnih podatkih

Fišer

Piasecki

Broda

2013

SLO2.0

Self Cite

View full text Add to dashboard Cite

Wordnet lahko izdelamo na podlagi že obstoječega tujejezičnega wordneta ali pa kot osnovo za gradnjo vzamemo korpusne podatke. Prvi pristop je preprostejši in enostavnejši, zaradi česar ga razvijalci tudi najpogosteje uporabljajo. Vendar ima ta pristop veliko pomanjkljivost, predvsem to, da tako izdelan vir ne odseva nujno jezika, za katerega je bil izdelan. Zato v pričujočem prispevku predstavljamo pristop, ki izhaja iz jezikovnih podatkov, pridobljenih iz jezikoslovno označenega referenčnega korpusa, iz katerega smo s pomočjo preprostih statističnih metod izluščili sezname semantično podobnih besed, ki smo jih nato vključili v wordnet za slovenščino. Pristop je bil prvotno razvit za poljščino, vendar je privlačen tudi za druge jezike, saj zanj potrebujemo minimalna jezikovnotehnološka orodja in vire, zato ga je enostavno uporabiti tudi za jezike, za katere obsežen wordnet ali podoben semantični leksikon še ne obstaja. Druga pomembna prednost uporabljenega pristopa pa je, da temelji na izpričani jezikovni rabi, pridobljeni iz korpusa, ki se nato kaže v jezikovno utemeljeni organizaciji besedišča v izdelani semantični mreži. Glede na to, da so vsi naši dosedanji pristopi za izdelovo slovenskega wordneta celotno strukturo prevzeli iz Princetonovega WordNeta, ki je bil izdelan za angleščino, bodo spodbudni rezultati, dobljeni s pričujočo metodo, koristno dopolnjevali obstoječo semantično mrežo.

show abstract

Metaheuristics for Tuning Model Parameters in Two Natural Language Processing Applications

Cited by 2 publications

References 6 publications

Paintball – Automated Wordnet Expansion Algorithm based on Distributional Semantics and Information Spreading

Paintball – Automated Wordnet Expansion Algorithm based on Distributional Semantics and Information Spreading

Utemeljevanje sloWNeta na korpusnih podatkih

Contact Info

Product

Resources

About