Apoorv Agarwal

Contextual Phrase-Level Polarity Analysis Using Lexical Affect Scoring and Syntactic N-Grams

We present a classifier to predict contextual polarity of subjective phrases in a sentence. Our approach features lexical scoring derived from the Dictionary of Affect in Language (DAL) and extended through WordNet, allowing us to automatically score the vast majority of words in our input avoiding the need for manual labeling. We augment lexical scoring with n-gram analysis to capture the effect of context. We combine DAL scores with syntactic constituents and then extract n-grams of constituents from all sentences. We also use the polarity of all syntactic constituents within the sentence as features. Our results show significant improvement over a majority class baseline as well as a more difficult baseline consisting of lexical n-grams.

Frame Semantic Tree Kernels for Social Network Extraction from Text

In this paper, we present work on extracting social networks from unstructured text. We introduce novel features derived from semantic annotations based on FrameNet. We also introduce novel semantic tree kernels that help us improve the performance of the best reported system on social event detection and classification by a statistically significant margin. We show results for combining the models for the two aforementioned subtasks into the overall task of social network extraction. We show that a combination of features from all three levels of abstractions (lexical, syntactic and semantic) are required to achieve the best performing system.

Predicting Interests of People on Online Social Networks

We introduce a new data set which contains both a self-declared friendship network and self-chosen attributes from a finite list defined by the social networking site. We propose Gaussian Field Harmonic Functions (GFHF), a state-of-the-art graph transduction algorithm, as a novel way of testing the relevance of the friendship network for predicting individual attributes. We show that the underlying self-declared friendship network allows us to predict some but not all attributes. We use Support Vector Machines (SVM) in conjunction with GFHF to show that other attributes such as age or languages spoken are also important.

Columbia NLP: Sentiment Detection of Sentences and Subjective Phrases in Social Media

We present two supervised sentiment detection systems which were used to compete in SemEval-2014 Task 9: Sentiment Analysis in Twitter. The first system (Rosenthal and McKeown, 2013) classifies the polarity of subjective phrases as positive, negative, or neutral. It is tailored towards online genres, specifically Twitter, through the inclusion of dictionaries developed to capture vocabulary used in online conversations (e.g., slang and emoticons) as well as stylistic features common to social media. The second system (Agarwal et al., 2011) classifies entire tweets as positive, negative, or neutral. It too includes dictionaries and stylistic features developed for social media, several of which are distinctive from those in the first system. We use both systems to participate in Subtasks A and B of SemEval2014 Task 9: Sentiment Analysis in Twitter. We participated for the first time in Subtask B: Message-Level Sentiment Detection by combining the two systems to achieve improved results compared to either system alone.

Parsing Screenplays for Extracting Social Networks from Movies

In this paper, we present a formalization of the task of parsing movie screenplays. While researchers have previously motivated the need for parsing movie screenplays, to the best of our knowledge, there is no work that has presented an evaluation for the task. Moreover, all the approaches in the literature thus far have been regular expression based. In this paper, we present an NLP and ML based approach to the task, and show that this approach outperforms the regular expression based approach by a large and statistically significant margin. One of the main challenges we faced early on was the absence of training and test data. We propose a methodology for using well structured screenplays to create training data for anticipated anomalies in the structure

Key Female Characters in Film Have More to Talk About Besides Men: Automating the Bechdel Test

The Bechdel test is a sequence of three questions designed to assess the presence of women in movies. Many believe that because women are seldom represented in film as strong leaders and thinkers, viewers associate weaker stereotypes with women. In this paper, we present a computational approach to automate the task of finding whether a movie passes or fails the Bechdel test. This allows us to study the key differences in language use and in the importance of roles of women in movies that pass the test versus the movies that fail the test. Our experiments confirm that in movies that fail the test, women are in fact portrayed as less-central and less-important characters.

Validating Literary Theories Using Automatic Social Network Extraction

In this paper, we investigate whether longstanding literary theories about nineteenthcentury British novels can be verified using computational techniques. Elson et al. (2010) previously introduced the task of computationally validating such theories, extracting conversational networks from literary texts. Revisiting their work, we conduct a closer reading of the theories themselves, present a revised and expanded set of hypotheses based on a divergent interpretation of the theories, and widen the scope of networks for validating this expanded set of hypotheses.

Labeling by landscaping: classifying tokens in context by pruning and decorating trees

State-of-the-art approaches to token labeling within text documents typically cast the problem either as a classification task, without using complex structural characteristics of the input, or as a sequential labeling task, carried out by a Conditional Random Field (CRF) classifier. Here we explore principled ways for structure to be brought to bear on the task. In line with recent trends in statistical learning of structured natural language input, we use a Support Vector Machine (SVM) classification framework deploying tree kernels. We then propose tree transformations and decorations, as a methodology for modeling complex linguistic phenomena in highly multi-dimensional feature spaces. We develop a general purpose tree engineering framework, which enables us to transcend the typically complex and laborious process of feature engineering. We build kernel based classifiers for two token labeling tasks: fine-grained event recognition, and lexical answer type detection in questions. For both, we show that in comparison with a corresponding linear kernel SVM, our method of using tree kernels improves recognition, thanks to appropriately engineering tree structures for use by the tree kernel. We also observe significant improvements when comparing with a CRF-based realization of structured prediction, itself performing at levels comparable to state-of-the-art.

Using Frame Semantics in Natural Language Processing

We summarize our experience using FrameNet in two rather different projects in natural language processing (NLP). We conclude that NLP can benefit from FrameNet in different ways, but we sketch some problems that need to be overcome.