Michele Vindigni

Corpus-Driven Unsupervised Learning of Verb Subcategorization Frames

The behavior of verbs in sublanguages is highly specific and does not follow general principles of lexical decomposition. NLP applications require specific lexicons for tasks like surface parsing and shallow semantic interpretation. The reduced set of verbal senses specific to a given domain is more appropriate for efficient processing in real world tasks (e.g. information extraction and retrieval). In this paper a method for learning verb subcategorization patterns from corpora is proposed. Conceptual clustering techniques are applied to the results of surface parsing in order to extract relevant domain typical senses and automatically build a lexicon of subcategorization frames. The aim is to learn a core of lexico-grammatical knowledge suitable to support more sophisticated parsing strategies to be applied in a target NLP application. Results derived for the Italian language from several corpora are presented.

AI/NLP technologies applied to spacecraft mission design

In this paper we propose the model of a prototypical NLP architecture of an information access system to support a team of experts in a scientific design task, in a shared and heterogeneous framework. Specifically, we believe AI/NLP can be helpful in several tasks, such as the extraction of implicit information needs enclosed in meeting minutes or other documents, analysis of explicit information needs expressed through Natural Language, processing and indexing of document collections, extraction of required information from documents, modeling of a common knowledge base, and, finally, identification of important concepts through the automatic extraction of terms. In particular, we envisioned this architecture in the specific and practical scenario of the Concurrent Design Facility (CDF) of the European Space Agency (ESA), in the framework of the SHUMI project (Support To HUman Machine Interaction) developed in collaboration with the ESA/ESTEC - ACT (Advanced Concept Team).

Agents Based Ontological Mediation in IE Systems

Building more adaptive SW applications is a crucial issue to scale up IE technology to the Web, where information is organized following different underlying knowledge and/or presentation models. Information agents are more and more being adopted to support extraction of relevant information from semi-structured web sources. To efficiently manage heterogeneous information sources they must be able to cooperate, to share their knowledge, and to agree upon appropriate terminology to be used during interaction. Being the internal knowledge representation possibly different for each participant, it reveals to be unfeasible to directly communicate concepts, while agents autonomy promotes abstraction from details about the internal structure of other agents. We will argue on main topics involved in adapting natural language to achieve semantic agreement in communication, and we will introduce a novel architecture based on a pool of intelligent agents. It will be done by defining a communication model that foresees a strong separation between terms and concepts, (being this difference often undervalued in the literature, where terms play the ambiguous roles of both concept labels and communication lexicon). For agents communicating through the language, lexical information embodies the possibility to “express” the underlying conceptualizations thus agreeing to a shared representation. To make the resulting architecture adaptive to the application domain three different agents typologies have been defined: resource agents, owning the target knowledge; service agents, providing basic skills to support complex activities and control agents, supplying the structural knowledge of the task, with coordination and control capabilities. We will focus on two dedicated service agents: a mediator, that will care about understanding the information an agent wants to express as well as the way to present it to others, and a translator, dealing with lexical misalignment due to different languages. The resulting agent community dynamically assumes the most appropriate configuration, in a transparent way with respect to the involved participants.

Demonstration of the CROSSMARC system

The EC-funded R&D project, CROSSMARC, is developing technology for extracting information from domain-specific web pages, employing language technology methods as well as machine learning methods in order to facilitate technology porting to new domains. CROSSMARC also employs localisation methodologies and user modelling techniques in order to provide the results of extraction in accordance with the users personal preferences and constraints. The systems implementation is based on a multi-agent architecture, which ensures a clear separation of responsibilities and provides the system with clear interfaces and robust and intelligent information processing capabilities.

Understanding the Web through its Language

A tight integration between ontological and linguistic knowledge is critical within the information processes of the Semantic Web. In Information Extraction, ontologies should include knowledge components neglected in domain conceptualizations generally used for other tasks. In this paper, we analyze such critical information in the light of existing applications. Accordingly, a methodology for semi-automatic development of an IE ontology integrating pre-existing domain and lexical knowledge is presented. The proposed ontological framework supports the discovery of new relations among known concepts by means of text processing, but also induction of new conceptual information.