Systematic comparison of neural architectures and training approaches for open information extraction
The goal of open information extraction (OIE) is to extract facts from natural language text, and to represent them as structured triples of the form (subject, predicate, object). For example, given the sentence »Beethoven composed the Ode to Joy.«, we are expected to extract the triple (Beethoven,...
| Main Authors: | , , , |
|---|---|
| Format: | Conference item |
| Language: | English |
| Published: |
Association for Computational Linguistics
2020
|
| _version_ | 1826302313694756864 |
|---|---|
| author | Hohenecker, P Mtumbuka, F Kocijan, V Lukasiewicz, T |
| author_facet | Hohenecker, P Mtumbuka, F Kocijan, V Lukasiewicz, T |
| author_sort | Hohenecker, P |
| collection | OXFORD |
| description | The goal of open information extraction (OIE) is to extract facts from natural language text, and to represent them as structured triples of
the form (subject, predicate, object). For example, given the sentence »Beethoven composed the Ode to Joy.«, we are expected to extract the triple (Beethoven, composed, Ode to Joy). In this work, we systematically compare different neural network architectures and training approaches, and improve the performance of the currently best models on the OIE16 benchmark (Stanovsky and Dagan, 2016) by 0.421 F1 score and 0.420 AUCPR, respectively, in our experiments (i.e., by more than 200% in both cases). Furthermore, we show that appropriate problem and loss formulations often affect the performance more than the network architecture. |
| first_indexed | 2024-03-07T05:45:43Z |
| format | Conference item |
| id | oxford-uuid:e7231759-fcd8-404e-9a8e-e79c789846a4 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T05:45:43Z |
| publishDate | 2020 |
| publisher | Association for Computational Linguistics |
| record_format | dspace |
| spelling | oxford-uuid:e7231759-fcd8-404e-9a8e-e79c789846a42022-03-27T10:36:27ZSystematic comparison of neural architectures and training approaches for open information extractionConference itemhttp://purl.org/coar/resource_type/c_5794uuid:e7231759-fcd8-404e-9a8e-e79c789846a4EnglishSymplectic ElementsAssociation for Computational Linguistics2020Hohenecker, PMtumbuka, FKocijan, VLukasiewicz, TThe goal of open information extraction (OIE) is to extract facts from natural language text, and to represent them as structured triples of the form (subject, predicate, object). For example, given the sentence »Beethoven composed the Ode to Joy.«, we are expected to extract the triple (Beethoven, composed, Ode to Joy). In this work, we systematically compare different neural network architectures and training approaches, and improve the performance of the currently best models on the OIE16 benchmark (Stanovsky and Dagan, 2016) by 0.421 F1 score and 0.420 AUCPR, respectively, in our experiments (i.e., by more than 200% in both cases). Furthermore, we show that appropriate problem and loss formulations often affect the performance more than the network architecture. |
| spellingShingle | Hohenecker, P Mtumbuka, F Kocijan, V Lukasiewicz, T Systematic comparison of neural architectures and training approaches for open information extraction |
| title | Systematic comparison of neural architectures and training approaches for open information extraction |
| title_full | Systematic comparison of neural architectures and training approaches for open information extraction |
| title_fullStr | Systematic comparison of neural architectures and training approaches for open information extraction |
| title_full_unstemmed | Systematic comparison of neural architectures and training approaches for open information extraction |
| title_short | Systematic comparison of neural architectures and training approaches for open information extraction |
| title_sort | systematic comparison of neural architectures and training approaches for open information extraction |
| work_keys_str_mv | AT hoheneckerp systematiccomparisonofneuralarchitecturesandtrainingapproachesforopeninformationextraction AT mtumbukaf systematiccomparisonofneuralarchitecturesandtrainingapproachesforopeninformationextraction AT kocijanv systematiccomparisonofneuralarchitecturesandtrainingapproachesforopeninformationextraction AT lukasiewiczt systematiccomparisonofneuralarchitecturesandtrainingapproachesforopeninformationextraction |