More Related Content
Viewers also liked (10)
Similar to Contextual Suggestion 2014 (20)
Recently uploaded (20)
Contextual Suggestion 2014
- 1. Better Contextual Suggestions from ClueWeb12 Using Domain Knowledge Inferred from The Open Web Thaer Samar Alejandro Bellogin and Arjen P. de Vries
- 2. Our Submission Contextual Suggestion model: 1. Find attractions in ClueWeb12 2. Generating user profiles 3. Similarity between candidate attractions and users 4. Rank suggestion per (user, context) pair RQ: Can we improve the performance of the contextual suggestions by applying domain knowledge? Approach: Filter collection using domain knowledge to create sub-collections Apply same contextual suggestion model to different sub-collections Compare differences in effectiveness Two runs based on two sub-collections created differently
- 3. Creating Sub-collections (1) GeoFiltered sub-collection Applying geographical filter Exact mention of the given contexts format: {City, ST} e.g., Miami, FL Exclude documents that mention multiple contexts e.g., a Wikipedia page about cities in Florida state
- 4. Creating Sub-collections (2) TouristFiltered sub-collection Applying domain knowledge extracted from the structure of the Open Web: Domain Oriented Manual list of tourist websites {yelp, tripadvisor, wikitravel, zagat, xpedia, orbitz, and travel.yahoo} From ClueWeb12 extract any document whose host in the list (TouristListFiltered) e.g., http://www.zagat.com/miami Expand TouristListFiltered Extract outlinks Search for outlinks in ClueWeb12 (TouristOutlinksFiltered)
- 5. TouristFiltered sub-collection Attraction Oriented Use Foursquare API to get attractions for given contexts If URL is missing for the attraction, then use Google API query: Cort辿s Restaurant Miami, FL For found attractions Get host names of their URLs (1,454 unique hosts) From ClueWeb12 get any document whose host from the above (AttractionFiltered) 50 attractions per context Format: attraction name, URL e.g., Cort辿s Restaurant, http://cortesrestaurant.com Miami, FL
- 6. Sub-collections Summary ClueWeb12 733,019,372 docs City, ST 8,883,068 docs TouristListFiltered (175,260) TouristOutlinksFiltered (97,678) AttractionsFiltered (102,604) GeoFiltered TouristFiltered
- 7. Generating User Profiles For each user: Aggregate descriptions of attractions rated by the user Split the aggregated descriptions into positive and negative profiles based on the ratings
- 8. Similarity Represent attractions and user profiles in weighted VSM Vector element <term, frequency> Cosine similarity
- 9. Ranked suggestions For each (user, context) pair Rank suggestions based on similarity score Generate titles to represent attraction: Extract from <title> or <header> tags Generate descriptions tailored to the user Extract content of <description> tag Break documents into sentences rank sentences based on their similarity with the user Concatenate until 512 bytes reached
- 10. Results (General Performance) TouristFiltered is better than the GeoFiltered
- 11. Analysis (TouristFiltered vs. GeoFiltered) Percentage of topics where TouristFiltered is better than equal to and worse than GeoFiltered e.g., ToursitFiltered gives better result for 33.1% of the judged topics
- 12. Analysis (decompose metrics dimensions ) P@5 and MRR consider three dimensions of relevance Geographical (geo), description (desc) and document (doc) relevance Considering the desc and doc relevance only The two runs have almost similar performance in the desc and doc dimensions
- 13. Analysis (decompose metrics evaluation ) Considering the geo aspect only TouristFiltered is more geographically appropriate
- 14. Analysis (Effect of sub-collection parts ) TouristFiltered sub-collection consists of three parts TouristListFiltered (TLF) TouristOutlinksFiltered (TOF) AttractionFiltered (AF) Measure how each part contributes to the performance Major improvement in the performance gained after adding the AttractionFiltered (AF) part
- 15. Conclusions Applying domain knowledge about sites that are more likely to offer attractions leads to better suggestions The best result are obtained when identifying attractions through specialized services like Foursquare Our approach emphasized the importance of accurate geo-information for high precision Reproducibility of research results: recommendations from Clueweb12
- 16. Future Work We can think of each part in TouristFiltered collection as a binary filter Each document in each part passes the corresponding filter Filter documents based on score obtained by: Combining different weighted filters Each filter can represent a different source of knowledge
- 17. Thanks!