While feeding multiple sentences to knews, we noticed an error. If we feed in, e.g. four sentences at once, or four sentences one after another, the output of 'all sentences' is quite similar (sometimes even exact similar) to the output of only the last sentence. The differences merely appear in the synsets (different sysnset ids, or no found synset) never in the structure.
However, if we replace the periods within the text (not the last one) with a semicolon, we get a more expected result, which is longer and holds information of all the four sentences. We do not yet know, however, if this comes with other downsides.
Also, we found it conspicuous, that in the 'all sentence semicolon' variant, an exzessiv amount of variables seems to be generated.
Look here for detailed in and output
The example is from https://cs.rochester.edu/nlp/rocstories/
Configuration:
[semantics] : boxer
[wsd], [el]: babelfy
[net]: wordnet
[boxer]: soap
While feeding multiple sentences to knews, we noticed an error. If we feed in, e.g. four sentences at once, or four sentences one after another, the output of 'all sentences' is quite similar (sometimes even exact similar) to the output of only the last sentence. The differences merely appear in the synsets (different sysnset ids, or no found synset) never in the structure.
However, if we replace the periods within the text (not the last one) with a semicolon, we get a more expected result, which is longer and holds information of all the four sentences. We do not yet know, however, if this comes with other downsides.
Also, we found it conspicuous, that in the 'all sentence semicolon' variant, an exzessiv amount of variables seems to be generated.
Look here for detailed in and output
The example is from https://cs.rochester.edu/nlp/rocstories/
Configuration:
[semantics] : boxer
[wsd], [el]: babelfy
[net]: wordnet
[boxer]: soap