5 years ago

Automatic Extraction of Examples for Word Sense Disambiguation

Automatic Extraction of Examples for Word Sense Disambiguation

Chapter 6

Chapter 6 Automatic Extraction of Examples for WSD The following Chapter is devoted particularly to our work and the description of the designed by us system. We will describe stepwise the process of word sense disambiguation that we employ starting with a short overview of the system (Section 6.1) and the data collection that we use - Section 6.2. A short discussion of how the data collection is preprocessed can be found in Section 6.3, followed by its division into training and test set in Section 6.4. We will look at the specific WSD algorithm that we selected (Section 6.5) as well as the parameter optimization which we considered (Section 6.6). In order to make it completely transparent how we acquired the results reported in Section 6.8 we will first present (Section 6.7) the scoring software that was provided by Senseval-3, which we employed for the scoring of our system. The final section of the chapter (Section 6.9) is an evaluation of the presented system. 6.1 Overview of the System The idea of how our system is constructed is pretty straightforward. As a semi-supervised one it is basically put together similarly to the one visualized in Figure 2.1 on page 20. The difference between a supervised WSD system and ours is delineated by the automatically labeled data that we add to our final collection of examples. It characterizes our approach as a semi-supervised WSD. This distinctness is depicted in Figure 6.1 on page 48 in the Data Collection component. We chose to exhibit this figure here, since it represents to a great extent the overview of our system. Later on in the section we will discuss in more detail each of its separate parts in order to give a deeper insight of our endeavors. How those several components function together is represented in Figure 6.1 on page 48. A very important part of the experimental process is the collection of data. We used the manually annotated data, but since our aim was to extend it we gathered examples from several online dictionaries and various corpora. The latter collection was to be automatically annotated 47

CHAPTER 6. AUTOMATIC EXTRACTION OF EXAMPLES FOR WSD 48 and together with the manually annotated one the final data collection was constructed (to be seen in the upper left corner of Figure 6.1). The next step was to preprocess the data and turn it into a training and test set needed for the training and testing procedures. For training, a specific WSD algorithm had to be selected together with the parameters that best fit the given task. After the word-expert is trained it is tested on the prepared test set and an evaluation of the results is attempted. At that point, if the evaluation is satisfying, the final word-expert/classifier is accepted. In case the results need to be improved a new start of the process at any of its states can be attempted. 6.2 Data Collection Figure 6.1: Our semi-supervised WSD system. Here we will describe in more detail the data we had collected for our experiments. We will pay attention mainly to the sources we used to gather the data, to the sense inventory according to which the senses were chosen and to the actual annotation procedure. 6.2.1 Sense Inventory The English lexical sample task (see Section 4.5.2), which served as the source for our manually annotated data used WordNet and Wordsmyth as sense inventories. The former was concerned with the possible senses for the nouns and the adjectives in the examples and the latter with

A Machine Learning Approach for Automatic Road Extraction - asprs
Selective Sampling for Example-based Word Sense Disambiguation
Word sense disambiguation with pattern learning and automatic ...
Word Sense Disambiguation Using Automatically Acquired Verbal ...
Word Sense Disambiguation The problem of WSD - PEOPLE
Using Machine Learning Algorithms for Word Sense Disambiguation ...
Performance Metrics for Word Sense Disambiguation
Word Sense Disambiguation - cs547pa1
Word Sense Disambiguation Using Selectional Restriction -
Using Lexicon Definitions and Internet to Disambiguate Word Senses
A Comparative Evaluation of Word Sense Disambiguation Algorithms
Semi-supervised Word Sense Disambiguation ... - ResearchGate
word sense disambiguation and recognizing textual entailment with ...
MRD-based Word Sense Disambiguation - the Association for ...
Word Sense Disambiguation: An Empirical Survey - International ...
Using unsupervised word sense disambiguation to ... - INESC-ID
Word Sense Disambiguation is Fundamentally Multidimensional
Using Meaning Aspects for Word Sense Disambiguation
Word-Sense Disambiguation for Machine Translation
Word Sense Disambiguation Using Association Rules: A Survey
KU: Word Sense Disambiguation by Substitution - Deniz Yuret's ...
Soft Word Sense Disambiguation
Similarity-based Word Sense Disambiguation
Word Sense Disambiguation with Pictures - CLAIR
NUS-ML: Improving Word Sense Disambiguation Using Topic ...
A Word-Sense Disambiguated Multilingual Wikipedia Corpus - UPC