Citation Context

...ifferent learning method; instead of using a single support vector machine classifier, it employed an ensemble of this classifier. This ensemble was implemented using the AdaBoostM1 algorithm in Weka =-=[3]-=-. Table 3 shows the evaluation results corresponding to our two submitted runs. It also shows (in the last row) the results for a 100% YES baseline (i.e., an answer validation system that validated al...

Citation Context

... “eight” to the example question, using as support text “…when eight animals parade by the principal street in Longyearbyen, a town of a thousand of inhabitants”. 1 This analysis was done by Freeling =-=[2]-=-, an open source suite of language analyzers. 2 It is not necessary to compute the term overlap between (T) and (H´) since it will be exactly the same.Table 2. Application of transformation patterns ...

Citation Context

... order to accomplish this recognition they have probed several approaches, ranging from simple ones taking advantage of lexical overlaps to more complexes founded on the use of a logic representation =-=[6]-=-. The approach based on lexical overlaps is quite simple, but surprisingly it has achieved very competitive results. Representative methods of this approach determine that H (the hypothesis) is entail...

Citation Context

... the other hand, the evaluation set for the Spanish AVE 2007 contains 564 answers (22.5% validated and 77.5% rejected) corresponding to 170 different questions. Details on these sets are described in =-=[7]-=-. 3 For the training process we considered all questions from the previous question answering CLEF campaigns.3.2 Results This section describes the experimental results of our participation at the AV...

Citation Context

... hypothesis) is entailed from T (the support text) only considering characteristics such as named entity overlaps [8], n-gram overlaps [4], as well as the size of the longest common subsequence (LCS) =-=[1,5]-=-. The simplicity is the strength of this approach, but also it is its weakness. For instance, [8] can easily recognize the textual entailment between “Lucy visits some friends” (H) and “Lucy goes with...

Citation Context

...ery competitive results. Representative methods of this approach determine that H (the hypothesis) is entailed from T (the support text) only considering characteristics such as named entity overlaps =-=[8]-=-, n-gram overlaps [4], as well as the size of the longest common subsequence (LCS) [1,5]. The simplicity is the strength of this approach, but also it is its weakness. For instance, [8] can easily rec...

Citation Context

...ses the calculus of the LCS. • It uses some manually constructed lexical patterns to help treating support texts containing an apposition phrase. This idea was taken from the COGEX logic-based system =-=[9]-=-. Its goal is to make explicit the relation between the two elements of the apposition phase. • Finally, it includes some new features denoting certain answer restrictions as imposed by the question’s...

Citation Context

...qa-accuracy. The results of Table 4 are not conclusive. However, it is interesting to comment that our QA system (that was the best one in 2005 and the second best one in 2006 in the Spanish QA task) =-=[10]-=- obtains a 35.88% of accuracy on the same questions set and ignoring evaluate the correct NIL questions. This fact indicates –in some way– that answer validation is useful, and that it could produce i...

Citation Context

...ts. Representative methods of this approach determine that H (the hypothesis) is entailed from T (the support text) only considering characteristics such as named entity overlaps [8], n-gram overlaps =-=[4]-=-, as well as the size of the longest common subsequence (LCS) [1,5]. The simplicity is the strength of this approach, but also it is its weakness. For instance, [8] can easily recognize the textual en...

Citation Context

... hypothesis) is entailed from T (the support text) only considering characteristics such as named entity overlaps [8], n-gram overlaps [4], as well as the size of the longest common subsequence (LCS) =-=[1,5]-=-. The simplicity is the strength of this approach, but also it is its weakness. For instance, [8] can easily recognize the textual entailment between “Lucy visits some friends” (H) and “Lucy goes with...