Kernel-based learning has been largely adopted in many semantic textual inference tasks. In particular, Tree Kernels (TKs) have been successfully applied in the modeling of syntactic similarity between linguistic instances in Question Answering or Information Extraction tasks. At the same time, lexical semantic information has been studied through the adoption of the so-called Distributional Semantics (DS) paradigm, where lexical vectors are acquired automatically from large-scale corpora. Recently, Compositional Semantics phenomena arising in complex linguistic structures have been studied in an extended paradigm called Distributional Compositional Semantics (DCS), where, for example, algebraic operators on lexical vectors have been defined to account for grammatically typed bi-grams or complex verb or noun phrases. In this paper, a novel kernel called Compositionally Smoothed Partial Tree Kernel is presented to integrate DCS operators into the tree kernel evaluation by also considering complex compositional nodes. Empirical results on well-known NLP tasks show that state-of-the-art performances can be achieved, without resorting to manual feature engineering, thus suggesting that a large set of Web and text mining tasks can be handled successfully by this kernel.

Kernel-based learning has been largely applied to semantic textual inference tasks. In particular, Tree Kernels (TKs) are crucial in the modeling of syntactic similarity between linguistic instances in Question Answering or Information Extraction tasks. At the same time, lexical semantic information has been studied through the adoption of the so-called Distributional Semantics (DS) paradigm, where lexical vectors are acquired automatically from large corpora. Notice how methods to account for compositional linguistic structures (e.g. grammatically typed bi-grams or complex verb or noun phrases) have been proposed recently by defining algebras on lexical vectors. The result is an extended paradigm called Distributional Compositional Semantics (DCS). Although lexical extensions have been already proposed to generalize TKs towards semantic phenomena (e.g. the predicate argument structures as for role labeling), currently studied TKs do not account for compositionality, in general. In this paper, a novel kernel called Compositionally Smoothed Partial Tree Kernel is proposed to integrate DCS operators into the tree kernel evaluation, by acting both over lexical leaves and non-terminal, i.e. complex compositional, nodes. The empirical results obtained on a Question Classification and Paraphrase Identification tasks show that state-of-the-art performances can be achieved, without resorting to manual feature engineering, thus suggesting that a large set of Web and text mining tasks can be handled successfully by the kernel proposed here.

English. Several textual inference tasks rely on kernel-based learning. In particular Tree Kernels (TKs) proved to be suitable to the modeling of syntactic and semantic similarity between linguistic instances. In order to generalize the meaning of linguistic phrases, Distributional Compositional Semantics (DCS) methods have been defined to compositionally combine the meaning of words in semantic spaces. However, TKs still do not account for compositionality. A novel kernel, i.e. the Compositional Tree Kernel, is presented integrating DCS operators in the TK estimation. The evaluation over Question Classification and Metaphor Detection shows the contribution of semantic compositions w.r.t. traditional TKs. Italiano. Sono numerosi i problemi di interpretazione del testo che beneficiano dall’applicazione di metodi di apprendimento automatico basato su funzioni kernel. In particolare, i Tree Kernel (TK) sono applicati alla modellazione di metriche di similarita sintattica e semantica tra espressioni linguistiche. Allo scopo di generalizzare i significati legati a sintagmi complessi, i metodi di Distributional Compositional Semantics combinano algebricamente i vettori associati agli elementi lessicali costituenti. Ad oggi i modelli di TK non esprimono criteri di composizionalita. In questo lavoro dimostriamo il beneficio di modelli di composizionalita applicati ai TK, in problemi di Question Classification e Metaphor Detection.

Empirical distributional methods account for the meaning of syntactic structures by combining word vectors according to algebraic operators. In this paper, a novel approach for semantic composition based on space projection techniques over lexical vector representations is proposed. In line with the principle of compositionality, the meaning of a phrase is modeled in terms of the subset of properties shared by co-occurring words. Syntactic bi-grams are thus projected in the so called Support Subspace, corresponding to such properties. State-of-the-art results are achieved in a well known phrase similarity task, used as a benchmark for this class of methods.

In this paper, an approach for semantic composition based on space projection operations over basic geometric lexical representations is proposed. Syntactic bi-grams are here projected in the so called Support Subspace, aimed at emphasizing the semantic features shared by the compound word. Empirical results are discussed over two tasks: a fine grained similarity estimation task over syntactically typed word pairs and the Semantic Textual Similarity (STS) taskas defined by SemEval 2012. The generalization capabilities ofthe proposed compositional operators are also investigated ina cross-linguistic scenario, i.e. over English and Italian.

Although distributional models of word meaning have been widely used in Information Retrieval achieving an effective representation and generalization schema of words in isolation, the composition of words in phrases or sentences is still a challenging task. Different methods have been proposed to account on syntactic structures to combine words in term of algebraic operators (e.g. tensor product) among vectors that represent lexical constituents. In this paper, a novel approach for semantic composition based on space projection techniques over the basic geometric lexical representations is proposed. In the geometric perspective here pursued, syntactic bi-grams are projected in the so called Support Subspace, aimed at emphasizing the semantic features shared by the compound words and better capturing phrase-specific aspects of the involved lexical meanings. State-of-the-art results are achieved in a well known benchmark for phrase similarity task and the generalization capability of the proposed operators is investigated in a cross-linguistic scenario, i.e. in the English and Italian Language.

This paper presents the UNITOR system that participated to the SemEval 2012 Task 6: Semantic Textual Similarity (STS). The task is here modeled as a Support Vector (SV) regression problem, where a similarity scoring function between text pairs is acquired from examples. The semantic relatedness between sentences is modeled in an unsupervised fashion through different similarity functions, each capturing a specific semantic aspect of the STS, e. g. syntactic vs. lexical or topical vs. paradigmatic similarity. The SV regressor effectively combines the different models, learning a scoring function that weights individual scores in a unique resulting STS. It provides a highly portable method as it does not depend on any manually built resource (e.g. WordNet) nor controlled, e. g. aligned, corpus.

Current Semantic Role Labeling technologies are based on inductive algorithms trained over large scale repositories of annotated examples. Frame-based systems currently make use of the FrameNet database but fail to show suitable generalization capabilities in out-of-domain scenarios. In this paper, a state-of-art system for frame-based SRL is extended through the encapsulation of a distributional model of semantic similarity. The resulting argument classification model promotes a simpler feature space that limits the potential overfitting effects. The large scale empirical study here discussed confirms that state-of-art accuracy can be obtained for out-of-domain evaluations.

Resources annotated with frame semantic information support the development of robust systems for shallow semantic parsing. Several researches proposed to automatically transfer the semantic information available for English corpora towards other resource-poor languages. In this paper, a semantic transfer approach is proposed based on Hidden Markov Models applied to aligned corpora. The experimental evaluation reported over an English-Italian corpus is successful, achieving 86% of accuracy on average, and improves on the state of the art methods for the same task.

The scenarios opened by the increasing availability, sharing and dissemination of music across the Web is pushing for fast, effective and abstract ways of organizing and retrieving music material. Automatic classification is a central activity to model most of these processes, thus its design plays a relevant role in advanced Music Information Retrieval. In this paper, we adopted a state-of-the-art machine learning algorithm, i.e. Support Vector Machines, to design an automatic classifier of music genres. In order to optimize classification accuracy, we implemented some already proposed features and engineered new ones to capture aspects of songs that have been neglected in previous studies. The classification results on two datasets suggest that our model based on very simple features reaches the state-of-art accuracy (on the ISMIR dataset) and very high performance on a music corpus collected locally.