Project Title: Determining the Most Efficient Methods By Which Computers Analyze Metaphor
BASIS Advisor: Ms. Bhattacharya
Internship Location: Computer History Museum
Onsite Mentor: Dr. Jelena Mitrovic
Artificial Intelligence is playing an increasingly large role in society, and has proven itself to be an extremely efficient tool in nearly every field that it is applied to. In order for AI to be useful, it must be accessible. For AI to be accessible, it needs to be able to understand the people who utilize it. My Senior Project specifically addresses human metaphor and how computers interpret such an abstract figure of speech. I’m conducting my research at the Computer History Museum, and I will be analyzing three different methods by which computers identify phrases as literal or metaphorical. The first method counts phrases as metaphors or not, based on how often a noun and its hyponyms/hypernyms occur with the adjective. The second compares adjectives and verbs to the paired noun, and if the noun is abstract compared to a concrete adjective, it is most likely a metaphor. The third uses a random-forest-classifier to create a data tree and select the most favorable branch. As of now, I don’t have a clear prediction for which method will be the fastest or most accurate, but I believe that the results will be very similar between methods. By finding the most accurate method, I feel that I can help increase the general ease with which computers understand humans.
My Posts
Week 12 – A Successful End
Success! After presenting my first presentation to my advisor and a class, I got some good feedback to improve my slides–I feel prepared for the upcoming Senior Project Committee presentations, between the new advice and additional practice. Beyond my presentation, however, I was actually able to bring my method’s accuracy up from a little over […]
Week 11 – Finishing Up
This past week was spent primarily working on my presentation. Compiling everything was relatively easy–I’ve been keeping a “journal” of sorts for my blog post drafts and project notes, and I referred to it extensively. Beyond my presentation, I’ve also been looking into different kinds of software to help me code the hypernym/hyponym method. At […]
Week 10 – The Final Stretch
There isn’t a whole lot I can write about this past week that’s relevant to my Senior Project, other than that this week was a welcome break from my Senior Project. I spent time in Paris, France with my family, and if nothing else, I feel refreshed and ready to finish the last leg of […]
Week 9 – Results!
Method 2, which analyzes word concreteness, is completely finished! I ran the method over 1,000 times and was able to achieve an average accuracy a little over 50%. To give a sense of context, previous algorithms written by professional teams with the same intent would achieve accuracies ranging from 60% to 85%. While I know […]
Week 8 – Success!
This week, to say the least, was productive. On Thursday, in addition to my usual weekly meeting, I was able to meet with Ms. Green, a language arts teacher at BASIS. Ms. Green had experience with NLP and was able to give me some good pointers on my project. She recommended that I use semantic […]
Week 7 – Hardships and Progress
This week, I began to code method 2, which looks at word concreteness and determines whether or not a phrase is metaphorical based on this information. I ran into some issues with method 1, because I need to find a database that can analyze word co-occurrence. It turns out that for AN phrases, method 1 […]
Week 6 – Code Beginnings
Week 6 was relatively uneventful, as I’m still in the very early phases of my code. I’ve been looking into how I should utilize the hypernyms and hyponyms in my code, and this will probably be the biggest issue I face with this method. After all, how far back should I look with the hypernym-hyponym […]
Week 5 – Random Forest Classifiers
Week 5 was spent reading the last of the research material, a paper titled “Metaphor Detection with Cross-Lingual Model Transfer.” The paper details Yulia Tsvetkov and her team’s research into metaphor detection, and how they used a random-forest classifier as their method. Tsvetkov addresses both Subject-Verb-Object (SVO) and Adjective-Noun (AN) metaphors, both of which I […]
Week 4: A Large Detour
As I’ve learned this week, research doesn’t always pan out as expected. My week 4 was spent reading up on the second method to be used in my project–the paper itself was interesting and offered yet another unique approach to phrase classification as either metaphorical or literal. The algorithm was devised by Peter Turney, with […]
Week 3 – Hypernyms and Hyponyms
This week, I started my research on the three different methods to be coded in my project. I read Xiaojin Zhu and Saisuresh Krishnakumaran’s “Hunting Elusive Metaphors Using Lexical Resources,” the first of the methods–beyond its relevance to my project, I found the paper to be a fantastic read! Zhu and Krishnakumaran’s algorithm is […]
Week 2: A Slight Detour
This past week was spent getting further acquainted with the subject matter of my project. I decided to read another piece of background research–the paper, titled “Distribution of Semantic Features Across Speech & Gesture by Humans and Machines,” is an analysis by MIT researchers Justine Cassell and Scott Prevost on how computers interact with people. […]
Week 1: Computers and Metaphors
Week 1: From nearly the very beginning of the Senior Project process, I had a pretty good idea of what I wanted to do my project on. Computer science has been my favorite subject for a long time, and it’s what I want to study when I attend college in the fall–it only seemed logical […]