Table 4

System performance metrics in pre-workshop evaluation

System performance measure (%)System output versus gold standard annotationSystem-assisted annotationsManual annotation
Textpresso
    Sentence level
        Category 4aSystem alone
            Recall37.9
            Precision77.5Curator 1bCurator 2b
            F-measure50.955.126.9
        Category 5aSystem alone41.763.3
            Recall39.747.537.8
            Precision81.5
            F-measure53.4
            GO annotation level
        Category 4aCurator 1Curator 2
            Recall37.114.5Curator 1bCurator 2b
            Precision78.377.886.839.5
            F-measure50.324.442.841.2
        Category 5aCurator 1Curator 257.340.3
            Recall32.211.3
            Precision75.071.4
            F-measure45.119.5
PCS
    Term-based EQscSystem aloneCurator 1Curator 2dCurator 3
        Recall65.047.038.050.0
        Precision60.057.065.067.0
        F-measure62.451.548.057.3
    Label-based EQscSystem aloneCurator 1curator 2dCurator 3
        Recall24.044.051.051.0
        Precision23.054.081.074.0
        F-measure23.548.562.660.4
Phenex + CharaparserPhenex
    Label-based EQscCurator 1Curator 2dCurator 3Curator 1Curator 2dCurator 3
        Recall51.038.066.037.063.036.0
        Precision58.070.084.049.088.060.0
        F-measure54.349.373.942.273.445.0
PubTator
    NLM indexing mention-levelSystem aloneCurator 1Curator 1
        Recall80.198.691.0
        Precision83.498.393.0
        F-measure81.798.092.0
    TAIR indexing document levelSystem aloneCurator 2Curator 2
        Recall76.090.091.0
        Precision73.977.175.0
        F-measure74.983.082.0
    TAIR triageSystem aloneCurator 2
        Recall68.684.6
        Precision80.5100.0
        F-measure74.192.0
PPInterFinder
    PPI algorithm aloneSystem aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.863.872.779.7
        Precision85.785.787.090.4
        F-measure76.973.279.284.7
    PPI algorithm (gene mention/ gene normalization)System aloneCurator 1Curator 2
        RecallNR46.946.9
        Precision85.785.7
        F-measure60.660.6
eFIP
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.288.289.577.8
        Precision94.779.085.070.0
        F-measure80.083.387.273.7
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR78.685.7100.077.8
        Precision91.785.783.377.8
        F-measure84.685.790.977.8
    Document-ranking
        nDCG93–100
T-HOD
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall70.056.022.024.042.0
        Precision79.532.026.040.042.0
        F-measure74.540.024.030.042.0
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall54.356.030.026.042.0
        Precision72.163.041.052.071.0
        F-measure62.059.035.035.053.0
System performance measure (%)System output versus gold standard annotationSystem-assisted annotationsManual annotation
Textpresso
    Sentence level
        Category 4aSystem alone
            Recall37.9
            Precision77.5Curator 1bCurator 2b
            F-measure50.955.126.9
        Category 5aSystem alone41.763.3
            Recall39.747.537.8
            Precision81.5
            F-measure53.4
            GO annotation level
        Category 4aCurator 1Curator 2
            Recall37.114.5Curator 1bCurator 2b
            Precision78.377.886.839.5
            F-measure50.324.442.841.2
        Category 5aCurator 1Curator 257.340.3
            Recall32.211.3
            Precision75.071.4
            F-measure45.119.5
PCS
    Term-based EQscSystem aloneCurator 1Curator 2dCurator 3
        Recall65.047.038.050.0
        Precision60.057.065.067.0
        F-measure62.451.548.057.3
    Label-based EQscSystem aloneCurator 1curator 2dCurator 3
        Recall24.044.051.051.0
        Precision23.054.081.074.0
        F-measure23.548.562.660.4
Phenex + CharaparserPhenex
    Label-based EQscCurator 1Curator 2dCurator 3Curator 1Curator 2dCurator 3
        Recall51.038.066.037.063.036.0
        Precision58.070.084.049.088.060.0
        F-measure54.349.373.942.273.445.0
PubTator
    NLM indexing mention-levelSystem aloneCurator 1Curator 1
        Recall80.198.691.0
        Precision83.498.393.0
        F-measure81.798.092.0
    TAIR indexing document levelSystem aloneCurator 2Curator 2
        Recall76.090.091.0
        Precision73.977.175.0
        F-measure74.983.082.0
    TAIR triageSystem aloneCurator 2
        Recall68.684.6
        Precision80.5100.0
        F-measure74.192.0
PPInterFinder
    PPI algorithm aloneSystem aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.863.872.779.7
        Precision85.785.787.090.4
        F-measure76.973.279.284.7
    PPI algorithm (gene mention/ gene normalization)System aloneCurator 1Curator 2
        RecallNR46.946.9
        Precision85.785.7
        F-measure60.660.6
eFIP
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.288.289.577.8
        Precision94.779.085.070.0
        F-measure80.083.387.273.7
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR78.685.7100.077.8
        Precision91.785.783.377.8
        F-measure84.685.790.977.8
    Document-ranking
        nDCG93–100
T-HOD
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall70.056.022.024.042.0
        Precision79.532.026.040.042.0
        F-measure74.540.024.030.042.0
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall54.356.030.026.042.0
        Precision72.163.041.052.071.0
        F-measure62.059.035.035.053.0

a4-Category search use ‘bag of words’ for (1) assay terms, (2) verbs, (3) cellular component terms, and (4) gene product names, whereas 5-Category search also include words for Table and Figures. bManual annotations don't necessarily correspond to either the 4- or 5-category search as curators do annotations for sentences that fit both criteria. cTerm-label EQs are entity-quality statements created strictly based on the original descriptions, independent of any ontologies, whereas the label-based EQs are the corresponding formal statements (using ontology terms). dCurator ignore an unspecified number of CharaParser proposals to save time.

Table 4

System performance metrics in pre-workshop evaluation

System performance measure (%)System output versus gold standard annotationSystem-assisted annotationsManual annotation
Textpresso
    Sentence level
        Category 4aSystem alone
            Recall37.9
            Precision77.5Curator 1bCurator 2b
            F-measure50.955.126.9
        Category 5aSystem alone41.763.3
            Recall39.747.537.8
            Precision81.5
            F-measure53.4
            GO annotation level
        Category 4aCurator 1Curator 2
            Recall37.114.5Curator 1bCurator 2b
            Precision78.377.886.839.5
            F-measure50.324.442.841.2
        Category 5aCurator 1Curator 257.340.3
            Recall32.211.3
            Precision75.071.4
            F-measure45.119.5
PCS
    Term-based EQscSystem aloneCurator 1Curator 2dCurator 3
        Recall65.047.038.050.0
        Precision60.057.065.067.0
        F-measure62.451.548.057.3
    Label-based EQscSystem aloneCurator 1curator 2dCurator 3
        Recall24.044.051.051.0
        Precision23.054.081.074.0
        F-measure23.548.562.660.4
Phenex + CharaparserPhenex
    Label-based EQscCurator 1Curator 2dCurator 3Curator 1Curator 2dCurator 3
        Recall51.038.066.037.063.036.0
        Precision58.070.084.049.088.060.0
        F-measure54.349.373.942.273.445.0
PubTator
    NLM indexing mention-levelSystem aloneCurator 1Curator 1
        Recall80.198.691.0
        Precision83.498.393.0
        F-measure81.798.092.0
    TAIR indexing document levelSystem aloneCurator 2Curator 2
        Recall76.090.091.0
        Precision73.977.175.0
        F-measure74.983.082.0
    TAIR triageSystem aloneCurator 2
        Recall68.684.6
        Precision80.5100.0
        F-measure74.192.0
PPInterFinder
    PPI algorithm aloneSystem aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.863.872.779.7
        Precision85.785.787.090.4
        F-measure76.973.279.284.7
    PPI algorithm (gene mention/ gene normalization)System aloneCurator 1Curator 2
        RecallNR46.946.9
        Precision85.785.7
        F-measure60.660.6
eFIP
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.288.289.577.8
        Precision94.779.085.070.0
        F-measure80.083.387.273.7
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR78.685.7100.077.8
        Precision91.785.783.377.8
        F-measure84.685.790.977.8
    Document-ranking
        nDCG93–100
T-HOD
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall70.056.022.024.042.0
        Precision79.532.026.040.042.0
        F-measure74.540.024.030.042.0
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall54.356.030.026.042.0
        Precision72.163.041.052.071.0
        F-measure62.059.035.035.053.0
System performance measure (%)System output versus gold standard annotationSystem-assisted annotationsManual annotation
Textpresso
    Sentence level
        Category 4aSystem alone
            Recall37.9
            Precision77.5Curator 1bCurator 2b
            F-measure50.955.126.9
        Category 5aSystem alone41.763.3
            Recall39.747.537.8
            Precision81.5
            F-measure53.4
            GO annotation level
        Category 4aCurator 1Curator 2
            Recall37.114.5Curator 1bCurator 2b
            Precision78.377.886.839.5
            F-measure50.324.442.841.2
        Category 5aCurator 1Curator 257.340.3
            Recall32.211.3
            Precision75.071.4
            F-measure45.119.5
PCS
    Term-based EQscSystem aloneCurator 1Curator 2dCurator 3
        Recall65.047.038.050.0
        Precision60.057.065.067.0
        F-measure62.451.548.057.3
    Label-based EQscSystem aloneCurator 1curator 2dCurator 3
        Recall24.044.051.051.0
        Precision23.054.081.074.0
        F-measure23.548.562.660.4
Phenex + CharaparserPhenex
    Label-based EQscCurator 1Curator 2dCurator 3Curator 1Curator 2dCurator 3
        Recall51.038.066.037.063.036.0
        Precision58.070.084.049.088.060.0
        F-measure54.349.373.942.273.445.0
PubTator
    NLM indexing mention-levelSystem aloneCurator 1Curator 1
        Recall80.198.691.0
        Precision83.498.393.0
        F-measure81.798.092.0
    TAIR indexing document levelSystem aloneCurator 2Curator 2
        Recall76.090.091.0
        Precision73.977.175.0
        F-measure74.983.082.0
    TAIR triageSystem aloneCurator 2
        Recall68.684.6
        Precision80.5100.0
        F-measure74.192.0
PPInterFinder
    PPI algorithm aloneSystem aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.863.872.779.7
        Precision85.785.787.090.4
        F-measure76.973.279.284.7
    PPI algorithm (gene mention/ gene normalization)System aloneCurator 1Curator 2
        RecallNR46.946.9
        Precision85.785.7
        F-measure60.660.6
eFIP
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR69.288.289.577.8
        Precision94.779.085.070.0
        F-measure80.083.387.273.7
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 1Curator 2
        RecallNR78.685.7100.077.8
        Precision91.785.783.377.8
        F-measure84.685.790.977.8
    Document-ranking
        nDCG93–100
T-HOD
    PMID-centric (sentence level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall70.056.022.024.042.0
        Precision79.532.026.040.042.0
        F-measure74.540.024.030.042.0
    Gene-centric (document level)System aloneCurator 1Curator 2Curator 3Curator 4
        Recall54.356.030.026.042.0
        Precision72.163.041.052.071.0
        F-measure62.059.035.035.053.0

a4-Category search use ‘bag of words’ for (1) assay terms, (2) verbs, (3) cellular component terms, and (4) gene product names, whereas 5-Category search also include words for Table and Figures. bManual annotations don't necessarily correspond to either the 4- or 5-category search as curators do annotations for sentences that fit both criteria. cTerm-label EQs are entity-quality statements created strictly based on the original descriptions, independent of any ontologies, whereas the label-based EQs are the corresponding formal statements (using ontology terms). dCurator ignore an unspecified number of CharaParser proposals to save time.

Close
This Feature Is Available To Subscribers Only

Sign In or Create an Account

Close

This PDF is available to Subscribers Only

View Article Abstract & Purchase Options

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

Close