Towards Keyphrase Assignment for Texts in Portuguese Language
Keyphrase assignment has often been confounded with keyphrase extraction, since the basic hypothesis is that a keyphrase of a text must be extracted from this text. Typically, keyphrase extraction approaches use a training set restricted to textual terms,
- PDF / 589,599 Bytes
- 12 Pages / 439.37 x 666.142 pts Page_size
- 14 Downloads / 159 Views
ract. Keyphrase assignment has often been confounded with keyphrase extraction, since the basic hypothesis is that a keyphrase of a text must be extracted from this text. Typically, keyphrase extraction approaches use a training set restricted to textual terms, reducing the learning capabilities of any inductive algorithm. Our research investigates ways to improve the accuracy of the keyphrase assignment systems for texts in Portuguese language by allowing classification algorithms to learn from non-textual terms as well. The basic assumption we have followed is that non-textual terms can be included into the training set by inference from an eventual semantic relationship with textual terms. In order to discover the latent relationship between non-textual and textual terms, we use deductive strategies to be applied in Portuguese common sense bases such as Wikipedia and InferenceNet. We show that algorithms that follow our approach outperform others that do not use the same methods introduced here. Keywords: Keyphrase extraction annotation Information retrieval
Keyphrase assignment
Semantic
1 Introduction The task of assigning a text with keyphrases is important because they enable text categorization [1], advertising [2], or simply for the purpose of summarizing the content to allow a rapid understanding of the subject matter [3]. This task, when done manually, is tedious and time consuming. When there is a need to consolidate a pre-defined vocabulary, this activity is non-trivial and its automation becomes mandatory. Traditionally, automatic keyphrase extraction concerns “the automatic selection of important and topical phrases from the body of a document” [4]. Its goal is to extract a set of phrases that are related to the main topics discussed in a given document [5]. In fact, the task of keyphrase assignment (discovery of keyphrases contained or no in the text) has often been confounded with keyphrase extraction, whose basic hypothesis is that a keyphrase of a text must be extracted from this text.
© Springer International Publishing Switzerland 2016 J. Silva et al. (Eds.): PROPOR 2016, LNAI 9727, pp. 165–176, 2016. DOI: 10.1007/978-3-319-41552-9_17
166
R. Silveira et al.
Our preliminary analysis from a corpus of news in Portuguese with keyphrases assigned by humans has shown that approximately 20 % of them are not in the text. Lately we have fortified the conclusions reached in the preliminary study by exploring a corpus of thesis and dissertations abstracts in Portuguese, which showed us that 55 % of the keyphrases assigned by the authors are not found in the text. The literature of automatic extraction of keyphrases is dominated by inductive learning (typically, classification). This kind of learning discovers patterns based on examples composed of statistical, structural and syntactic features of textual terms such as their frequency, their topological position in the text, and external resource-based features computed based on information gathered from resources other, such as knowledge bases (e
Data Loading...
