|
|
|
| [Introduction] |
[Description] |
[System] |
[Contact us] |
[Videos] |
[References] |
|
OASYS is a client-server system. The server
builds a database of opinions, which are available to users when requested through a
browser.
Server
The server is composed of four crawlers, which retrieve documents from American, Pakistani,
Spanish and Italian sources. For the first two the language is English whereas the other two have their own language (Spanish and Italian). For these documents we use commercial translators from foreign languages to English in order to run our algorithms (written for English language).
From each document retrieved, we save the date and a list of relevant topics. Then the opinion analysis algorithms are run for each topic and for each document, and the opinions result found are stored in the database. OASYS has been running since September 2005; so far our database has retrieved
3,061,066 documents.
Client
It is possible to access upon password request and query OASYS at http://oasys.umiacs.umd.edu/oasys/oasys.php.
From this page users can manage the crawlers by stopping them or adding,
deleting or modifyng sources to crawl. To query OASYS, users chose the sources, time interval and list of topics that define the opinion search. A graph of opinion intensity over time is displayed as the result of the search. A more detailed description of the results is displayed when individual data points on the graph are clicked.
Figure 1 is a screendump of the query interface. A simple query is performed by typing a desired name for the topic for which we want to read the opinions from the sources and click on the Oasys Search button. It is possible to obtain general global results from each different country (America, Spain, Italy and Pakistan) by using the by Country link or obtain local opinions for each source by using the by Sources link.
The menu on the right allows to execute advanced queries (change interval time, choose which sources select for quering, see statistical opinions by day, week and month on some fixed topics, compare two different graphs for our fixed topics and see the how the results change in presence of intensifiers).
|
|
Figure 1
|
Figure 2 is the resulting graph for the topic Berlusconi run by using the by Sources view. (This query has been executed on May 12 2006). As you can notice, the topic Berlusconi is reported by different news sources, three of which are italians (http://www.corriere.it, http://www.ilfoglio.it, http://www.repubblica.it). It is possible to click on each of those points to get the list of all the documents retrieved where Berlusconi appears as a topic. Each list starts with the precise day when the documents have been extracted. For each document there are both the quantitative and qualitative scores computed by our algorithms. It is also possible to retrieve and read that document by simply clicking on it.
|
|
Figure 2
|
|
|
| Developed at UMIACS (University of Maryland) - Webmaster: Diego Reforgiato Recupero |
