Victor R. Martinez

Sentiment Characterization of an Urban Environment via Twitter

We propose a statistical study of sentiment produced in an urban environment by collecting tweets submitted in a certain timeframe. Each tweet was processed using our own sentiment classifier and assigned either a positive or a negative label. By calculating the average mood, we were able to run a Mann-Withney’s U test to evaluate differences in the calculated mood per day of week. We found that all days of the week had significantly different medians. We also found positive correlations between Mondays and the rest of the week.

Linguistic analysis of differences in portrayal of movie characters

A computer implemented method for analyzing media content includes a step of providing a plurality of narrative files formatted in human readable format. Each narrative file includes a script and/or dialogues tagged with character names along with auxiliary information. Each script includes a plurality of portrayals performed by an associated actor or character. Linguistic representations of content of the narrative files in both abstract and semantic forms is determined. The linguistic representations are connected to higher order representations and mental states. The linguistic representations are connected to behavior and action. Interplay between language constructs and demographics of content creators is analyzed. Content representations towards individuals/groups are adapted to reflect heterogeneity in preferences.

The ELISA Situation Frame extraction for low resource languages pipeline for LoReHLT’2016

This paper describes the Situation Frame extraction pipeline developed by team ELISA as a part of the DARPA Low Resource Languages for Emergent Incidents program. Situation Frames are structures describing humanitarian needs, including the type of need and the location affected by it. Situation Frames need to be extracted from text or speech audio in a low resource scenario where little data, including no annotated data, are available for the target language. Our Situation Frame pipeline is the final step of the overall ELISA processing pipeline and accepts as inputs the outputs of the ELISA machine translation and named entity recognition components. The inputs are processed by a combination of neural networks to detect the types of needs mentioned in each document and a second post-processing step connects needs to locations. The resulting Situation Frame system was used during the first yearly evaluation on extracting Situation Frames from text, producing encouraging results and was later successfully adapted to the speech audio version of the same task.

Multimodal Representation of Advertisements Using Segment-level Autoencoders

Automatic analysis of advertisements (ads) poses an interesting problem for learning multimodal representations. A promising direction of research is the development of deep neural network autoencoders to obtain inter-modal and intra-modal representations. In this work, we propose a system to obtain segment-level unimodal and joint representations. These features are concatenated, and then averaged across the duration of an ad to obtain a single multimodal representation. The autoencoders are trained using segments generated by time-aligning frames between the audio and video modalities with forward and backward context. In order to assess the multimodal representations, we consider the tasks of classifying an ad as funny or exciting in a publicly available dataset of 2,720 ads. For this purpose we train the segment-level autoencoders on a larger, unlabeled dataset of 9,740 ads, agnostic of the test set. Our experiments show that: 1) the multimodal representations outperform joint and unimodal representations, 2) the different representations we learn are complementary to each other, and 3) the segment-level multimodal representations perform better than classical autoencoders and cross-modal representations -- within the context of the two classification tasks. We obtain an improvement of about 5% in classification accuracy compared to a competitive baseline.

Social Influence and Emotional State While Shopping

New technologies are opening novel ways to help people in their decision-making while shopping. From crowd-generated customer reviews to geo-based recommendations, the information to make the decision could come from different social circles with varied degrees of expertise and knowledge. Such differences affect how much influence the information has on the shopping decisions. In this work, we aim to identify how social influence when it is mediated by modern and ubiquitous communication (such as that provided by smartphones) can affect people’s shopping experience and especially their emotions while shopping. Our results showed that large amount of information affects emotional state in costumers, which can be measured in their physiological response. Based on our results, we conclude that integrating smartphone technologies with biometric sensors can create new models of customer experience based on the emotional effects of social influence while shopping.

Semi-Supervised Approach to Named Entity Recognition in Spanish Applied to a Real-World Conversational System

In this paper, we improve the named-entity recognition NER capabilities for an already existing text-based dialog system TDS in Spanish. Our solution is twofold: first, we developed a hidden Markov model part-of-speech POS tagger trained with the frequencies from over 120-million words; second, we obtained 2,i¾?283 real-world conversations from the interactions between users and a TDS. All interactions occurred through a natural-language text-based chat interface. The TDS was designed to help users decide which product from a well-defined catalog best suited their needs. The conversations were manually tagged using the classical Penn Treebank tag set, with the addition of an ENTITY tag for all words relating to a brand or product. The proposed system uses an hybrid approach to NER: first it looks up each word in a previously defined catalog. If the word is not found, then it uses the tagger to tag it with its appropriate POS tag. When tested on an independent conversation set, our solution presented a higher accuracy and higher recall rates compared to a current development from the industry.

Spatial-temporal characterization of an urban environment using Twitter

This paper presents a study of the usage of Twitter within the context of urban activity. We retrieved a set of tweets submitted by users located in Mexico City. Tweets were labeled as either positive or negative mood using a sentiment analyzer implementation. By calculating the average mood, we were able to run a Mann-Withneys U test to evaluate differences in the calculated mood per day of week. We found that all days of the week had significantly different medians with Sunday being the most positive day and Thursday the most negative. Additionally, we study the location for the tweets as an indicator important events and landmarks around the city.

Erratum: Sentiment Characterization of an Urban Environment via Twitter

We propose a statistical study of sentiment produced in an urban environment by collecting tweets submitted in a certain timeframe. Each tweet was processed using our own sentiment classifier and assigned either a positive or a negative label. By calculating the average mood, we were able to run a Mann-Withney’s U test to evaluate differences in the calculated mood per day of week. We found that all days of the week had significantly different medians. We also found positive correlations between Mondays and the rest of the week.

weRemember: letting ad patients to enjoy their home and their families

weRemember was designed to provide elderly people suffering from Alzheimers disease (AD) a relative independence at home and a new way to communicate and interact with their family. Our solution offers support for AD patients helping them to longer deal with the disease while living at home with their family instead of moving into a nursing home. Following an iterative design approach, a number of prototypes were evaluated with potential users and their feedback was used to enhance the family experience. During the prototype evaluation we found that the system could have a positive impact both on the relationship between the patient and the caregivers as well as on the patient home experience.