August 15, 2013 in News
Edward Roy and Leo Xuzhang Lin presented their summer research at the campus-wide UMass Amherst REU Poster Session on Friday, August 2. While Roy worked with Jeffrey McPherson to look at improving the usability of current semantic information systems, Lin worked with Dr. Tony McCaffrey on a project about automatic information retrieval from patents for design innovation. Their project abstracts as well as their posters can be seen below.
Improving Usability of Semantic Information
This project focuses on improving existing user interfaces of information management software for application in industry using an existing ontology interface, OntoWiki. Ontologies are information structures that are organized in a class based system with properties linking information together. In the past, the Center for e-Design at UMass Amherst along with other institutions has developed a collection of ontologies called the e-Design Framework. This collection of ontologies can be used for modeling, sharing, and integrating engineering design knowledge. Industry can utilize the e-Design Framework to store and organize vast quantities of information that is transparent through any step of the design process. The current problem with deploying the e-Design Framework is the lack of a user-friendly interface. OntoWiki has the capabilities to interface with ontologies, such as the e-Design Framework. However, the default OntoWiki interface currently lacks the necessary features to display ontologies in a user-friendly manner. The tree structure of the ontology is hidden when it is displayed in the OntoWiki user interface. All details of the structure exist and the class definitions remain intact, but no visual representation of the structure is available to the user. Information like this is paramount when a non-experienced user is attempting to locate a specific instance inside the ontology. Users can also locate instances by querying the ontology. This can be done using the built –in “Filter” extension, but this extension does not currently afford the capabilities necessary for the e-Design Framework. In order to fix these problems I have added a visual hierarchy tree system, repaired and improved the filtering capabilities, and improved the overall usability and functionality of OntoWiki in order to allow non-experienced users to access ontologies. These new features must be tested in order to measure the improvement from the original Further improvements on the appearance of OntoWiki can further extend the ontology using community by allowing people with no prior ontology knowledge to access ontologies.
Automatic Information Retrieval from Patents for Design Innovation
Leo Xuzhang Lin
Current patent classifications are complex and do not accurately represent the information content of the patent. Many classifications are too general or specific and moreover none of them represents the content of the patent. This increases the amount of time needed to search for a specific patent and decreases the accuracy of search. Improving patent search will not only enable innovation by finding and building on pre-existing patents, but also improve problem solving by finding accurate solutions residing in patents.
There had been extensive previous works by Hirtz et al that defined engineering design’s function by its energy flow and set of verbs describing the function. My research focuses on researching and creating an information retrieval algorithm that can determine a patent’s energy distribution and identify its functional verbs (defined by Hirtz et al) to supplement the limited metadata from the patent database. The algorithm I had created evaluates the energy distribution and calculates the frequency of the various functional verbs. The energy types are represented by the percentage of their occurrence in the patent (ex. 15% Mechanical, 85% Electrical). Using this information, we can get a general understanding of what the patent does through its functional verbs and what type of energy is used to achieve the solution.
The extracted information will be used to supplement the Analogy Finder search engine created by Dr. Anthony McCaffrey. It searches for analogous patents by its goal and synonyms of that goal. However, the search returns a large pool of results that requires further refinement. The energy type data will be used as individual limiting parameters so the user can look at patents that use a specific energy. Furthermore, it can be used to calculate the content similarity between two patents allowing ordering by uniqueness.