Catálogo de publicaciones - libros

One of the main differences between Spoken Document Retrieval (SDR) systems and Text Retrieval systems is the need of a segmentation process that detects the story boundaries. However, until now, SDR researchers have not paid attention in building correct segments more than considering slidding windows of a fixed size in time. In this paper, new methodology for evaluating segments to SDR task, and the evaluation of three possible strategies are presented over the TREC-9 SDR collection. Moreover, the justification of each strategy is discussed.

We define the Generalised Similarity Measure (GSM) as a means of uniformly and efficiently storing linguistic information to search for answers in Question Answering (QA) systems. It computes the similarity between a question representation and those of possible answers in a document collection as a database query. Linguistic knowledge from different sources can be used and combined in the GSM, allowing to find matches even with imperfect representations. To show the viability of the concept, we have implemented the GSM in a proof-of-concept QA system for German, employing information from WordNet and FrameNet. First experiments have been promising, large-scale tests are underway.

The paper presents MDDQL as a query language suitable for multi-lingual conceptual querying of collections of databases from a graphical user interface or from an application programming one. The query language, however, has been specified and implemented with the parametric theory of linguistic diversity in mind such that syntactic and semantic parsing of multi-lingual query expressions becomes quite simple and guarantees identical query results regardless the preferred natural language. We present a parsing algorithm, which shows that it is quite easy to formulate a query regardless the underlying type order of a natural language, be it , , or , etc., and still being able to grasp the meaning of the query at a minimal computational effort possible.

Much currently transmitted information takes the form of e-mails or SMS text messages and so extracting information from such short messages is increasingly important. The words in a message can be partitioned into the syntactic structure, terms from the domain of discourse and the data being transmitted. This paper describes a light-weight Information Extraction component which uses pattern matching to separate the three aspects: the structure is supplied as a template; domain terms are the metadata of a data source (or their synonyms), and data is extracted as those words matching placeholders in the templates.

Software requirements specification is a critical activity of the software process, as errors at this stage inevitably lead to problems later on in system design and implementation. The requirements are written in natural language, with the potential for ambiguity, contradiction or misunderstanding, or simply an inability of developers to deal with a large amount of information. This paper proposes a methodology for the natural language processing of textual descriptions of the requirements of an unlimited natural language and their automatic mapping to the object-oriented analysis model.

In this paper, the description of an image in natural language is carried out. The main idea is that from an image, with objects without movement, it is possible to obtain phrases in Spanish describing the position among the objects. In order to put this description into effect, we place ourselves in a theoretical model in which a cognitive-semantic analysis of linguistic units such as the prepositions (on), (in), (between) and the verb (to touch) is realized. This analysis will allow to establish rules which will determine the relationship or position among the objects.

For most English words, dictionaries give various senses: e.g., “”can stand for a financial institution, shore, set, etc. Automatic selection of the sense intended in a given text has crucial importance in many applications of text processing, such as information retrieval or machine translation: e.g., “() ” is to be translated into Spanish as “() ” whereas “() ()” as “() ().” To choose the optimal combination of the intended senses of all words, Lesk suggested to consider the global coherence of the text, i.e., which we mean the average relatedness between the chosen senses for all words in the text. Due to high dimensionality of the search space, heuristics are to be used to find a near-optimal configuration. In this paper, we discuss several such heuristics that differ in terms of complexity and quality of the results. In particular, we introduce a dimensionality reduction algorithm that reduces the complexity of computationally expensive approaches such as genetic algorithms.