Steve Welch

Semantic Lexicon Expansion for Concept-Based Aspect-Aware Sentiment Analysis

We have developed a prototype for sentiment analysis that is able to identify aspects of an entity being reviewed, along with the sentiment polarity associated to those aspects. Our approach relies on a core ontology of the task, augmented by a workbench for bootstrapping, expanding and maintaining semantic assets that are useful for a number of text analytics tasks. The workbench has the ability to start from classes and instances defined in an ontology and expand their corresponding lexical realizations according to target corpora. In this paper we present results from applying the resulting semantic asset to enhance information extraction techniques for concept-level sentiment analysis. Our prototype(Demo at http://bit.ly/1svngDi) is able to perform SemSA’s Elementary Task (Polarity Detection), Advanced Task #1 (Aspect-Based Sentiment Analysis), and Advanced Task #3 (Topic Spotting).

Mining Relations from Unstructured Content.

Extracting relations from unstructured Web content is a challenging task and for any new relation a significant effort is required to design, train and tune the extraction models. In this work, we investigate how to obtain suitable results for relation extraction with modest human efforts, relying on a dynamic active learning approach. We propose a method to reliably generate high quality training/test data for relation extraction - for any generic user-demonstrated relation, starting from a few user provided examples and extracting valuable samples from unstructured and unlabeled Web content. To this extent we propose a strategy which learns how to identify the best order to human-annotate data, maximizing learning performance early in the process. We demonstrate the viability of the approach (i) against state of the art datasets for relation extraction as well as (ii) a real case study identifying text expressing a causal relation between a drug and an adverse reaction from user generated Web content.

Sonora: A Prescriptive Model for Message Authoring on Twitter

Within social networks, certain messages propagate with more ease or attract more attention than others. This effect can be a consequence of several factors, such as topic of the message, number of followers, real-time relevance, person who is sending the message etc. Only one of these factors is within a user’s reach at authoring time: how to phrase the message. In this paper we examine how word choice contributes to the propagation of a message.

User-Centric Ontology Population

Ontologies are a basic tool to formalize and share knowledge. However, very often the conceptualization of a specific domain depends on the particular user’s needs. We propose a methodology to perform user-centric ontology population that efficiently includes human-in-the-loop at each step. Given the existence of suitable target ontologies, our methodology supports the alignment of concepts in the user’s conceptualization with concepts of the target ontologies, using a novel hierarchical classification approach. Our methodology also helps the user to build, alter and grow their initial conceptualization, exploiting both the target ontologies and new facts extracted from unstructured data. We evaluate our approach on a real-world example in the healthcare domain, in which adverse phrases for drug reactions, as extracted from user blogs, are aligned with MedDRA concepts. The evaluation shows that our approach has high efficacy in assisting the user to both build the initial ontology (\({{\mathrm{{\textit{HITS}\,@10}}}}\) up to 99.5%) and to maintain it (\({{\mathrm{{\textit{HITS}\,@10}}}}\) up to 99.1%).

Exploring the Efficiency of Batch Active Learning for Human-in-the-Loop Relation Extraction.

Domain-specific relation extraction requires training data for supervised learning models, and thus, significant labeling effort. Distant supervision is often leveraged for creating large annotated corpora however these methods require handling the inherent noise. On the other hand, active learning approaches can reduce the annotation cost by selecting the most beneficial examples to label in order to learn a good model. The choice of examples can be performed sequentially, i.e. select one example in each iteration, or in batches, i.e. select a set of examples in each iteration. The optimization of the batch size is a practical problem faced in every real-world application of active learning, however it is often treated as a parameter decided in advance. In this work, we study the trade-off between model performance, the number of requested labels in a batch and the time spent in each round for real-time, domain specific relation extraction. Our results show that the use of an appropriate batch size produces competitive performance, even compared to a fully sequential strategy, while reducing the training time dramatically.

Multi-lingual Concept Extraction with Linked Data and Human-in-the-Loop

Ontologies are dynamic artifacts that evolve both in structure and content. Keeping them up-to-date is a very expensive and critical operation for any application relying on semantic Web technologies. In this paper we focus on evolving the content of an ontology by extracting relevant instances of ontological concepts from text. We propose a novel technique which is (i) completely language independent, (ii) combines statistical methods with human-in-the-loop and (iii) exploits Linked Data as bootstrapping source. Our experiments on a publicly available medical corpus and on a Twitter dataset show that the proposed solution achieves comparable performances regardless of language, domain and style of text. Given that the method relies on a human-in-the-loop, our results can be safely fed directly back into Linked Data resources.