Article Structure
Abstract
Conventional Automated Essay Scoring (AES) measures may cause severe problems when directly applied in scoring Automatic Speech Recognition (ASR) transcription as they are error sensitive and unsuitable for the characteristic of ASR transcription.
Introduction
The assessment of learners’ language abilities is a significant part in language learning.
Related Work
Conventional AES systems usually exploit textual features to assess the quality of writing mainly in three different facets: the content facet, the arrangement facet and the language usage facet.
Data
As characteristics of the data determine the effectiveness of our methods, the details of it will be introduced first.
Approach
The approach used in this paper is to build a standard FST for the current examination topic.
Experiments
In this section, the proposed features and our FST methods will be evaluated on the corpus we mentioned above.
Conclusion and Future work
The aforementioned experiments indicate three points.
Topics
edit distance
- Here, we apply the edit distance to measure how best the path is.Page 5, “Approach”
- This means the best path is the word sequence path in the FST which has the smallest edit distance compared with the to-be-scored transcription’s word sequences .Page 5, “Approach”
- EDcost is the edit distance from the transcription to the paths which start at state 0 and end at the endPage 5, “Approach”
- (3) The minEDcostG) is the accumulated minimum edit distance from state 0 to state j, and the cost(i,j) is the cost of insertion, deletion or substitution from state j to state i.Page 6, “Approach”
- ,Xp_1} in phase p. In phrase p, we compute the best path and its edit distance from the transcription for all the to-be-calculated states which is the Xp shown in Figure 4.Page 6, “Approach”
- After computing all the phrases, the best path and its edit distances of the end states are obtained.Page 6, “Approach”
- Then the final best path is the one with the smallest edit distance .Page 6, “Approach”
- o The Edit Distance (ED):Page 6, “Approach”
- The edit distance is the linear combination of the weights of insertion, deletion and substitution.Page 6, “Approach”
- The Normalized Edit Distance (NED):Page 6, “Approach”
- The extension of the FST model is mainly reflected in calculation of the edit distance of the best path.Page 7, “Approach”
See all papers in Proc. ACL 2012 that mention edit distance.
See all papers in Proc. ACL that mention edit distance.
Back to top.
Latent Semantic
- Compared with the Latent Semantic Analysis with Support Vector Regression (LSA-SVR) method (stands for the conventional measures), our FST method shows better performance especially towards the ASR transcription.Page 1, “Abstract”
- In the LSA-SVR method, each essay transcription is represented by a latent semantic space vector, which is regarded as the features in the SVR model.Page 3, “Related Work”
- The LSA (Deerwester et al., 1990) considers the relations between the dimensions in conventional vector space model (VSM) (Salton et al., 1975), and it can order the importance of each dimension in the Latent Semantic Space (LS S).Page 3, “Related Work”
See all papers in Proc. ACL 2012 that mention Latent Semantic.
See all papers in Proc. ACL that mention Latent Semantic.
Back to top.
