On the time-dependent occupancy distribution of the g/g/1 queuing system
Probability in the Engineering and Informational Sciences
LCC-SRN: LCC's SRN system for SemEval 2007 task 4
SemEval '07 Proceedings of the 4th International Workshop on Semantic Evaluations
| Hi-index | 0.00 |
In text, there are different types of semantic relations between various constituents. The automatic acquisition of semantic relations from text is an important problem in natural language processing since it represents the foundation for understanding text semantics. In this dissertation, we focus only on the relations between nouns; i.e. the relations encoded by genitives, noun compounds, adjectival phrases, and others. Two different approaches for the detection of semantic relations between nouns in open-text were used: (1) a per-relation approach, that focuses on detecting one relation at a time; and (2) a per-pattern approach, that tries to detect all the semantic relations encoded by a lexico-syntactic pattern. The per-relation method was used for the detection of part-whole and possession relations. The recognition of the semantic relation was done by checking a set of semantic constraints that are found using a new supervised machine learning algorithm, called Iterative Semantic Specialization. The semantic constraints are determined using the semantic information of the nouns (their meaning and WordNet information). The Iterative Semantic Specialization algorithm yielded an 82.05% f-measure for the part-whole relation and 62.37% f-measure for the possession relation. In contrast, the per-pattern approach introduces an algorithm called Semantic Scattering which detects a WordNet boundary that allows us to discriminate among the relations with a higher probability. The recognition consists of identifying the best value from the boundary and assigning the best relation corresponding to that value. The method is applied to two of the most common noun phrase lexico-syntactic patterns: the genitives and the noun compounds. This approach provides a 79.85% f-measure for of-genitives (the genitives of the form “NP of NP”), 78.75% f-measure for s-genitives (the genitives of the form “NP's NP”), and 61.58% f-measure for noun compounds.