Please enable JavaScript.
Coggle requires JavaScript to display documents.
Measuring Short Text Semantic Equivalence Using Multiple Similarity…
Measuring Short Text Semantic Equivalence Using Multiple Similarity Measurements
Zhu and Lan, 2013
Features
-
String similarity
Weighted Word Overlap (WWO) (Saric et al., 2012)
Longest common sequence (LCS) (Allison and Dix, 1986)
word n-grams similarity (Lyon et al., 2001)
Corpus based similarity
Latent Semantic Analysis (LSA) (Landauer et al.,1997)
Co-occurrence Retrieval Model (CRM) (Weeds, 2003)
-
Number Similarity
we adopt two features following (Saric ˇ
et al., 2012), which are computed as follow... (p. 127)
-
-
Results
Core task
-
-
On MSRpar dataset, we
can see that the corpus-based measure achieves the
best, then the knowledge-based measure and the MT
measure follow.
Room for improvement
-
For the core task, in our
future work we will consider the measures to evaluate the sentence difference as well.