• Aucun résultat trouvé

Measuring the Effect of Discourse Structure on Sentiment Analysis

N/A
N/A
Protected

Academic year: 2021

Partager "Measuring the Effect of Discourse Structure on Sentiment Analysis"

Copied!
14
0
0

Texte intégral

Loading

Figure

Table 1. Quantitative overview of the annotated data (in percents) SE SN SI O SEandSI + – both no polarity M R 50 2 29 14 5 45.48 33.78 4 16.74 N R 22 6 49 2 12 17.40 55 4 23.60
Fig. 1. Two examples of produced discourse annotation
Table 3. Overall polarity ratings in both corpus genres
Table 4. Overall multi-scale ratings in both corpus genres

Références

Documents relatifs

The work presented here is part of a larger project, the ANNODIS project, the aims of which is to create an annotated corpus of French written texts for the study of

(iv); they may actually effect the transition from one major discourse unit to another by encapsulating the essential discourse content of the preceding unit

In every model, each CNN layer used 64 filters with kernel size 5, each CNN layer was followed by a max-pooling layer with pool size 4, the last pool layer was followed by a

Table 1 also dis- tinguishes di ff erent types of annotated data with a breakdown by approach: on the one hand there are segmented elementary discourse units (EDU), rhetorical

init, describing the word or expression which ac- tivated the analysis, tgt, describing the part of the sentence which contains information related to the target of the opinion,

From the point of view of the dependency patterns (the inputs of the neural network), a procedure for exploring model changes [19] would consist of: i) specify a time frame of a

Les sections termes et définitions correspondent à la définition d’un thésaurus donnée par la norme ISO 25964–1:2011 : “vocabulaire contrôlé et structuré

The notion of sentence - as it is defined in syntactic, semantic, graphic and prosodic terms - is not a suitable maximal unit for the prosodic and syntactic annotation of