• Training the data: In order for the buzzword or keyword to be recognized by the system, it must first be trained. For this training, we used the parts of speech method. We tagged couple resumes and tested it for recognition but we were getting many errors. For example, the main problem would be the key skill. Whenever we entered a skill like java, it would recognize it in other places in the resume instead of only in skills. This was a problem and would often give us wrong resumes. I realized that the problem was that there is a different format for most resumes, but most headings would be the same. I noticed a similar problem in my pattern recognition course which I took during my spring 2015 semester. Of course, the application was completely different but I remember my professor telling me that more training would always result in more accurate results. So I implemented this by training more than 100 resumes, while my other team members tried to find other solutions. I noticed the improvement in recognition and we were able to achieve more success by training more data.
• Connecting Hadoop application to Web UI: I faced a problem initially while connecting hadoop application to the web UI. In the web UI, when changes were done, they were not being reflecting from the hadoop application. When …show more content…
The table that gets created often had null values or inaccurate values. The main issue was that when submitting to the server, it wouldn’t give any response and that was same with the hive server, where the table gets stored and where it needs to be retrieved from. After struggling with this error for a while and researching about it, I learned that there different drivers for hive. So after finding the appropriate drivers and getting the permissions from master to slave for the hive server connection, this issue was