Generic Framework for Knowledge Classification-1
1 like284 views
This document summarizes a presentation about developing a generic framework for classifying knowledge in an answer hub. It describes developing an algorithm that achieved 72% accuracy in classifying questions. It discusses challenges like sparse data and large number of features that were addressed using chi-square testing. Skills gained include machine learning, NLP, Lucene, and productizing research. The work aligned with the presenter's interest in text and machine learning.
Generic Framework for Knowledge Classification-1
- 2. 2 Generic Framework For Knowledge Classification By Venkata Vineel
- 3. 3 Agenda • Introduction • Problem at Hand • How is it solved ? • Challenges • Skills and Career alignment • Q & A
- 4. 4 Introduction • Masters in Computer Science University of Utah, SaltLakeCity, UT • Systems Engineering Intern Internal tools team - Knowledge Management Interests: Scalability challenges, Machine Learning and Visualization.
- 5. 5 Problem at Hand • Generic Framework for classifying knowledge • Classifying questions in Answer Hub
- 6. 6 How did I solve ?? • Developed an generic algorithm. • Answer Hub Knowledge Base that learns.
- 7. 7 Project High Points • 72 % percent accuracy has been achieved. 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 1 3 5 7 9 11 13 15 17 19 21 23 Rank Statastics No of Questions RANK CATEGORIES
- 8. 8 Confusion matrix Categories V3 GBX C3 Hadoop BES DAL Raptor Stratus Security Pla>orm General User Tracking ExperimentaEon Service Frameworks Search Services Sherlock Batch Frameword Trinity Commerce OS Teradata AnalyEcs Pla>orm Total V3 1552 2 1 2 6 263 217 3 23 455 2 41 290 9 3 6 0 0 0 0 2875 GBX 1 68 0 0 0 6 37 0 1 9 1 26 4 8 0 0 0 1 0 0 162 C3 0 0 318 1 1 25 27 54 5 32 1 6 1 4 0 1 0 1 1 0 478 Hadoop 0 0 2 173 1 10 8 0 0 20 1 3 4 0 3 0 0 0 0 0 225 BES 11 0 0 0 300 59 39 1 0 5 0 1 22 0 0 0 0 0 0 0 438 DAL 67 0 1 0 3 2307 89 0 2 16 0 13 99 5 0 1 0 0 0 0 2603 Raptor 11 10 5 2 25 396 5352 3 62 212 26 184 337 25 6 17 0 0 1 0 6674 Stratus 1 0 82 2 1 40 188 435 4 40 0 13 6 0 2 1 0 1 0 0 816 Security Pla>orm 4 0 0 0 0 32 38 0 174 11 0 6 129 1 0 1 0 0 0 0 396 General 100 2 12 15 6 129 258 16 13 1200 3 88 64 29 4 3 0 0 5 0 1947 User Tracking 3 0 0 1 0 16 43 0 3 8 126 41 10 1 0 0 0 0 0 0 252 ExperimentaEon 1 1 0 0 0 27 40 0 1 8 0 868 29 1 0 0 0 0 3 0 979 Service Frameworks 124 3 0 0 6 90 299 2 67 83 0 56 1977 38 5 3 0 11 0 0 2764 Search Services 0 1 1 0 1 5 9 1 2 8 0 4 32 163 0 0 0 0 0 0 227 Sherlock 2 0 0 4 0 67 31 2 0 17 0 29 19 0 85 0 0 0 0 0 256 Batch Frameword 11 0 0 2 2 100 92 2 2 10 0 2 22 0 0 67 0 0 1 0 313 Trinity 0 0 0 0 0 0 0 0 0 0 0 4 1 1 0 0 0 0 0 0 6 Commerce OS 0 0 0 0 0 10 48 0 4 15 0 14 15 8 0 0 0 103 0 0 217 Teradata 0 0 1 1 0 10 0 0 0 0 1 16 2 1 0 1 0 0 49 0 82 AnalyEcs Pla>orm 0 0 1 1 0 5 1 0 1 23 1 14 0 3 1 0 0 0 1 11 63 Total 1888 87 424 204 352 3597 6816 519 364 2172 162 1429 3063 297 109 101 0 117 61 11 21773 Percentage correct 82.20339 78.16092 75 84.80392 85.22727 64.13678 78.52113 83.81503 47.8021978 55.24862 77.77777778 60.74177747 64.54456415 54.88215488 77.98165 66.33663366 #DIV/0! 88.03418803 80.32787 100
- 9. 9 Challenges and How Did We Overcome Those • Sparse data. • Large number of features. • Chi- Square test came to the rescue.
- 10. 10 Skills Obtained • Lucene • Literature survey of existing techniques • Machine Learning and NLP • Exposure to productizing research
- 11. 11 Alignment With My Career Path • Interested in Text and Machine Learning. • eBay has tonnes of data.
- 12. 12 Future Scope for Improvement • User profile • Support Vector Machine, TF-IDF and k-NN algorithms
- 13. 13 Q&A