Archive for October, 2015
October 30th, 2015, by Tim Finin, posted in NLP, NLP, Semantic Web
In this week’s ebiquity lab meeting (10:30am Monday Nov 2), Tim Finin will describe recent work on the Kelvin information extraction system and its performance in two tasks in the 2015 NIST Text Analysis Conference. Kelvin has been under development at the JHU Human Language Center of Excellence for several years. Kelvin reads documents in several languages and extracts entities and relations between them. This year it was used for the Coldstart Knowledge Base Population and Trilingual Entity Discovery and Linking tasks. Key components in the tasks are a system for cross-document coreference and another that links entities to entries in the Freebase knowledge base.
October 29th, 2015, by Tim Finin, posted in Machine Learning, NLP, RDF, Semantic Web
Lyrics Augmented Multi-modal
1:00pm Friday 30 October, ITE 325b
In an increasingly mobile and connected world, digital music consumption has rapidly increased. More recently, faster and cheaper mobile bandwidth has given the average mobile user the potential to access large troves of music through streaming services like Spotify and Google Music that boast catalogs with tens of millions of songs. At this scale, effective music recommendation is critical for music discovery and personalized user experience.
Recommenders that rely on collaborative information suffer from two major problems: the long tail problem, which is induced by popularity bias, and the cold start problem caused by new items with no data. In such cases, they fall back on content to compute similarity. For music, content based features can be divided into acoustic and textual domains. Acoustic features are extracted from the audio signal while textual features come from song metadata, lyrical content, collaborative tags and associated web text.
Research in content based music similarity has largely been focused in the acoustic domain while text based features have been limited to metadata, tags and shallow methods for web text and lyrics. Song lyrics house information about the sentiment and topic of a song that cannot be easily extracted from the audio. Past work has shown that even shallow lyrical features improved audio-only features and in some tasks like mood classification, outperformed audio-only features. In addition, lyrics are also easily available which make them a valuable resource and warrant a deeper analysis.
The goal of this research is to fill the lyrical gap in existing music recommender systems. The first step is to build algorithms to extract and represent the meaning and emotion contained in the song’s lyrics. The next step is to effectively combine lyrical features with acoustic and collaborative information to build a multi-modal recommendation engine.
For this work, the genre is restricted to Rap because it is a lyrics-centric genre and techniques built for Rap can be generalized to other genres. It was also the highest streamed genre in 2014, accounting for 28.5% of all music streamed. Rap lyrics are scraped from dedicated lyrics websites like ohhla.com and genius.com while the semantic knowledge base comprising artists, albums and song metadata come from the MusicBrainz project. Acoustic features are directly used from EchoNest while collaborative information like tags, plays, co-plays etc. come from Last.fm.
Preliminary work involved extraction of compositional style features like rhyme patterns and density, vocabulary size, simile and profanity usage from over 10,000 songs by over 150 artists. These features are available for users to browse and explore through interactive visualizations on Rapalytics.com. Song semantics were represented using off-the-shelf neural language based vector models (doc2vec). Future work will involve building novel language models for lyrics and latent representations for attributes that is driven by collaborative information for multi-modal recommendation.
Committee: Drs. Tim Finin (Chair), Anupam Joshi, Pranam Kolari (WalmartLabs), Cynthia Matuszek and Tim Oates
October 25th, 2015, by Tim Finin, posted in Pervasive Computing, Security
In this week’s ebiquity meeting (10:30am Monday, 26 October 2015 in ITE346 at UMBC), Sandeep Nair will talk about his research on securing the cyber-physical systems in modern vehicles.
Vehicles changed from being just mechanical devices which will just obey the commands to a smarter Sensor-ECU-Actuator systems which sense the surroundings and take necessary smart actions. A modern car has around forty to hundred different ECU’s, possibly communicating, to make intelligent decisions. But recently, there is a lot of buzz in the research community on hacking and taking control of vehicles. These literature describe and document the different ways to take control of vehicles. In this talk, we will first discuss what makes this kind of hacking possible? Then we will continue with different logical ways to do this and discuss some proposed mechanisms to protect it. We then propose a context aware mechanism which can detect these unsafe behaviors in the vehicle and describe the challenges associated with them.
October 24th, 2015, by Tim Finin, posted in AI, NLP
Abhay Kashyap, Lushan Han, Roberto Yus, Jennifer Sleeman, Taneeya Satyapanich, Sunil Gandhi and Tim Finin, Robust Semantic Text Similarity Using LSA, Machine Learning and Linguistic Resources, Language Resources and Evaluation, Springer, to appear.
Semantic textual similarity is a measure of the degree of semantic equivalence between two pieces of text. We describe the SemSim system and its performance in the *SEM~2013~and SemEval-2014~tasks on semantic textual similarity. At the core of our system lies a robust distributional word similarity component that combines Latent Semantic Analysis and machine learning augmented with data from several linguistic resources. We used a simple term alignment algorithm to handle longer pieces of text. Additional wrappers and resources were used to handle task specific challenges that include processing Spanish text, comparing text sequences of different lengths, handling informal words and phrases, and matching words with sense definitions. In the *SEM~2013~task on Semantic Textual Similarity, our best performing system ranked first among the 89 submitted runs. In the SemEval-2014~task on Multilingual Semantic Textual Similarity, we ranked a close second in both the English and Spanish subtasks. In the SemEval-2014~task on Cross–Level Semantic Similarity, we ranked first in Sentence–Phrase, Phrase-Word, and Word-Sense subtasks and second in the Paragraph-Sentence subtask.
October 16th, 2015, by Tim Finin, posted in Big data, Machine Learning, NLP, NLP
Demystifying Word2Vec – A Hands-on Tutorial
10:30am Monday, 19 October 2015 **ITE 456**
In the world of NLP, Word2Vec is one of the coolest kids in town! But what exactly is it and how does it work? More importantly, how is it used/useful?
For the first 10-15 minutes, we will go over distributional an distributed representation of words and the neural language model behind Word2Vec. We will also briefly look at doc2vec, the extension of Word2Vec for longer pieces of text.
For the remainder of the time (45-60 minutes), we will get our feet wet by running Word2Vec on a dataset which will then be followed by discussions about potential ways it can be useful for your own work.
What to bring – Any computing machine with Python installed, lots of curiosity and some delicious snacks for me maybe? We will use the excellent gensim package for python to run Word2Vec along with cython to speed things up. If you aren’t familiar with Python or don’t like it, no worries! It’s really just 5-6 lines of code! The training dataset will be provided. If you wish to bring your own, that’s cool too.
NOTE: We will hold this week’s Ebiquity meeting in ITE 456.