206 research outputs found
Using Semantic Models for Robust Natural Language Human Robot Interaction
No full textWhile robotic platforms are moving from industrial to consumer applications, the need of flexible and intuitive interfaces becomes more critical and the capability of governing the variability of human language a strict requirement. Grounding of lexical expressions, i.e. mapping words of a user utterance to the perceived entities of a robot operational scenario, is particularly critical. Usually, grounding proceeds by learning how to associate objects categorized in discrete classes (e.g. routes or sets of visual patterns) to linguistic expressions. In this work, we discuss how lexical mapping functions that integrate Distributional Semantics representations and phonetic metrics can be adopted to robustly automate the grounding of language expressions into the robotic semantic maps of a house environment. In this way, the pairing between words and objects into a semantic map facilitates the grounding without the need of an explicit categorization. Comparative measures demonstrate the viability of the proposed approach and the achievable robustness, quite crucial in operational robotic settings
A context-based model for sentiment analysis in twitter
No full textMost of the recent literature on Sentiment Analysis over Twitter is tied to the idea that the sentiment is a function of an incoming tweet. However, tweets are filtered through streams of posts, so that a wider context, e.g. A topic, is always available. In this work, the contribution of this contextual information is investigated. We modeled the polarity detection problem as a sequential classification task over streams of tweets. A Markovian formulation of the Support Vector Machine discriminative model as embodied by the SVMhmm algorithm has been here employed to assign the sentiment polarity to entire sequences. The experimental evaluation proves that sequential tagging effectively embodies evidence about the contexts and is able to reach a relative increment in detection accuracy of around 20% in F1 measure. These results are particularly interesting as the approach is flexible and does not require manually coded resources
Context-aware spoken language understanding for human robot interaction
No full textL’interpretazione di comandi
espressi nei confronti di piattaforme
robotiche e un processo strettamente `
legato al contesto operativo in cui avviene
l’interazione. In questo lavoro, presentiamo
LU4R - adaptive spoken Language
Understanding 4 Robots, un sistema per
l’elaborazione automatica di comandi vocali,
dipendente dall’ambiente in cui il comando
viene espresso. Il sistema proposto,
implementato come una cascata
di passi di elaborazione semantica, e`
stato progettato seguendo un’architettura
Client/Server, per ridurre i requisiti di integrazione
con le piattaforme robotiche esistenti.Robots operate in specific environments and the correct interpretation of linguistic interactions depends on physical, cognitive and language-dependent aspects triggered by the environment. In this work, we present LU4R - adaptive spoken Language Understanding 4 Robots, a Spoken Language Understanding chain for the semantic interpretation of robotic commands, that is sensitive to the operational environment. The system has been designed according to a Client/Server architecture in order to be easily integrated with the vast plethora of robotic platforms
Using semantic maps for robust natural language interaction with robots
No full textModern robotic architectures are equipped with sensors enabling a deep analysis of the environment. In this work, we aim at demonstrating that such perceptual information (here modeled through semantic maps) can be effectively used to enhance the language understanding capabilities of the robot. A robust lexical mapping function based on the Distributional Semantics paradigm is here proposed as a basic model of grounding language towards the environment. We show that making such information available to the underlying language understanding algorithms improves the accuracy throughout the entire interpretation process
Context-aware Models for Twitter Sentiment Analysis
No full textRecent works on Sentiment Analysis over Twitter are tied to the idea that the sentiment can
be completely captured after reading an incoming tweet. However, tweets are filtered through
streams of posts, so that a wider context, e.g. a topic, is always available. In this work, the
contribution of this contextual information is investigated for the detection of the polarity of
tweet messages. We modeled the polarity detection problem as a sequential classification task over
streams of tweets. A Markovian formulation of the Support Vector Machine discriminative model
has been here adopted to assign the sentiment polarity to entire sequences. The experimental
evaluation proves that sequential tagging better embodies evidence about the contexts and is
able to increase the accuracy of the resulting polarity detection process. These evidences are
strengthened as experiments are successfully carried out over two different languages: Italian
and English. Results are particularly interesting as the approach is flexible and does not rely on
any manually coded resources
LU4R: adaptive spoken language understanding for robots
No full textService robots are expected to operate in specific environments, where the presence of humansplays a key role. It is thus essential to enable for a natural and effective communication amonghumans and robots. One of the main features of such robotics platforms is the ability to react tospoken commands. This requires a comprehensive understanding of the user utterance to triggerthe robot reaction. Moreover, the correct interpretation of linguistic interactions depends onphysical, cognitive and language-dependent aspects related to the environment. In this work, wepresent the latest version of LU4R - adaptive spoken Language Understanding 4 Robots, a Spo-ken Language Understanding framework for the semantic interpretation of robotic commands,that is sensitive to the operational environment. The overall system is designed according to aClient/Server architecture in order to be easily deployed in a vast plethora of robotic platforms.Moreover, an improved version of HuRIC - Human-Robot Interaction Corpus is presented. Themain novelty presented in this paper is the extension to commands expressed in Italian. In orderto prove the effectiveness of such system, we also present some empirical results in both Englishand Italian computed over the new HuRIC resource
GQA-it: Italian Question Answering on Image Scene Graphs
No full textThe recent breakthroughs in the field of deep learning have lead to state-of-the-art results in several Computer Vision and Natural Language Processing tasks such as Visual Question Answering (VQA). Nevertheless, the training requirements in cross-linguistic settings are not completely satisfying at the moment. The datasets suitable for training VQA systems for non English languages are still not available, thus representing a significant barrier for most neural methods. This paper explores the possibility of acquiring in a semiautomatic fashion a large-scale dataset for VQA in Italian. It consists of more than 1 M question-answer pairs over 80k images, with a test set of 3,000 question-answer pairs manually validated. To the best of our knowledge, the models trained on this dataset represent the first attempt to approach VQA in Italian, with experimental results comparable with those obtained on the English original material
Robust Spoken Language Understanding for House Service Robots
No full textService robotics has been growing significantly in thelast years, leading to several research results and to a numberof consumer products. One of the essential features of theserobotic platforms is represented by the ability of interactingwith users through natural language. Spoken commands canbe processed by a Spoken Language Understanding chain, inorder to obtain the desired behavior of the robot. The entrypoint of such a process is represented by an Automatic SpeechRecognition (ASR) module, that provides a list of transcriptionsfor a given spoken utterance. Although several well-performingASR engines are available off-the-shelf, they operate in a generalpurpose setting. Hence, they may be not well suited in therecognition of utterances given to robots in specific domains. Inthis work, we propose a practical yet robust strategy to re-ranklists of transcriptions. This approach improves the quality of ASRsystems in situated scenarios, i.e., the transcription of roboticcommands. The proposed method relies upon evidences derivedby a semantic grammar with semantic actions, designed tomodel typical commands expressed in scenarios that are specificto human service robotics. The outcomes obtained throughan experimental evaluation show that the approach is able toeffectively outperform the ASR baseline, obtained by selectingthe first transcription suggested by the AS
Actionable ethics through neural learning
No full textWhile AI is going to produce a great impact on society, its alignment with human values and expectations is an essential step towards a correct harnessing of AI potentials for good. There is a corresponding growing need for mature and established technical standards to enable the assessment of an AI application as the evaluation of its graded adherence to formalized ethics. This is clearly dependent on methods to inject ethical awareness at all stages of an AI application development and use. For this reason we introduce the notion of Embedding Principles of ethics by Design (EPbD) as a comprehensive inductive framework. Although extending generic AI applications, it mainly aims at learning the ethical behaviour through numerical optimization, i.e. deep neural models. The core idea is to support ethics by integrating automated reasoning over formal knowledge and induction from ethically enriched training data. A deep neural network is proposed here to model both the functional as well as the ethical conditions characterizing a target decision. In this way, the discovery of latent ethical knowledge is enabled and made available to the learning process. The application of the above framework to a banking application, i.e. AI-driven Digital Lending, is used to show how accurate classification can be achieved without neglecting the ethical dimension. Results over existing datasets demonstrate that the ethical compliance of the sources can be used to output models able to optimally fine tune the balance between business and ethical accuracy
A markovian kernel-based approach for Italian speech acT labEliNg
No full textThis paper describes the UNITOR system that participated to the itaLIan Speech acT labEliNg task within the context of EvalIta 2018. A Structured Kernel-based Support Vector Machine has been here applied to make the classification of the dialogue turns sensitive to the syntactic and semantic information of each utterance, without relying on any task-specific manual feature engineering. Moreover, a specific Markovian formulation of the SVM is adopted, so that the labeling of each utterance depends on speech acts assigned to the previous turns. The UNITOR system ranked first in the competition, suggesting that the combination of the adopted structured kernel and the Markovian modeling is beneficial
- …
