TEST.txt

Bad Queries
1. cs 121 - first 2 are relevant but the next 2 are not and then 5th one is relevant
2. software engineering - all irrelevant hits
3. iftekhar ahmed - only last 2 are relevant
4. master of software engineering - first 5 links are all from the same owner, processing time > 300 ms
5. research uci computer science 2001 - slow queries
6. what is software testing and how much - slow queries
7. search engine - irrelevant hits only last 1
8. What is the importance of information systems in the modern world - slow, 3/5 are irrelevant
9. Who is the greatest in the world - slow
10. the best way to predict the future is to invent it - slow

Good Queries
1. master of human-computer interaction
2. machine learning
3. ACM
4. ics undergraduate student affairs office in donald bren hall
5. alex thornton
6. mswe
7. phd in computer science
8. informatics is everywhere
9. artificial intelligence
10. research

To improve ranking performance we implemented cosine similarity and tf-idf scoring. We realized that when
we treated them separately we were not always getting the best searches so, we multiplied the cosine similarity
score with the average tf-idf score. This made the searches more relevant.

To improve runtime performance we created 3 partial indexes. In these partial indexes we tokenized them and
used multithreading to save index processing time. These indexes included, the token, the url's/doc number, and the tf scores
These partial indexes where then converted by pickle where they were turned into binary. We also ensured O(1) look up time when we could.