Module 9: Additional Patents Deep Dive
50+ New PatentsComprehensive collection of Google patents discovered through extended research that complement the core 65 patents.
Trust & Authority Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 10,268,641 | Search result ranking based on trust | CSE TrustRank - uses Custom Search Engine annotations |
| EP1817697A2 | Link-based spam detection | TrustRank methodology for separating spam from good pages |
| US 8,458,196B1 | System and method for determining topic authority | Topical authority scoring based on document topics |
Key Concept: Trust-Based Ranking
Google's trust ranking differs from Yahoo's TrustRank. Google uses:
- Distance from trusted seed sites
- Custom Search Engine annotations from experts
- Knowledge-based trust (factual accuracy)
Freshness & QDF Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,832,088B1 | Freshness-based ranking | Query Deserves Freshness (QDF) scoring |
QDF Algorithm
The freshness engine determines when queries need fresh results by:
- Monitoring news site activity on topics
- Counting user selections of news results
- Computing fresh-seeking query values
When QDF Activates:
- Breaking news events
- Trending topics
- Seasonal/recurring events
- Hot topics with sudden interest spikes
Crawling & Indexing Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,042,112B1 | Scheduler for search engine crawler | Segment-based crawler scheduling |
| US 8,666,964B1 | Managing items in crawl schedule | Recrawl scheduling with due dates |
| US 20120078874A1 | Search Engine Indexing | Caffeine-era indexing architecture |
Crawl Priority Factors
Local Search Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,046,371B2 | Scoring local search results based on location prominence | Local prominence scoring |
| US 8,312,010B1 | Local business ranking using mapping information | Map-based business ranking |
| US 7,752,210B2 | Determining geographical location from IP address | IP-based geolocation |
Local Ranking Factors (Patent-Based)
- Relevance - Category and keyword match
- Distance - Proximity to searcher/specified location
- Prominence - Reviews, citations, link signals
Duplicate Detection Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 7,734,627B1 | Document similarity detection | Shingling for near-duplicate detection |
| US 7,779,002B1 | Detecting query-specific duplicate documents | Query-relevant duplicate detection |
| US 8,015,162B2 | Detecting duplicate and near-duplicate files | Multi-pass duplicate identification |
| US 10,223,406B2 | Entity normalization via name normalization | Name standardization for entity deduplication |
Shingling Process
Documents are compared using "shingles" (overlapping word sequences):
- Create shingles from document text
- Hash shingles for efficient comparison
- Calculate Jaccard similarity
- Cluster near-duplicates
Query Understanding Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,055,669B1 | Search queries improved based on query semantic information | Hummingbird foundations |
| US 8,577,907B1 | Search queries improved based on query semantic information | Semantic query improvement |
| US 7,636,714B1 | Determining query term synonyms within query context | Context-aware synonym detection |
| US 20020147578A1 | Query reformulation | Term expansion and synonyms |
Query Reformulation Pipeline
Visual & Image Search Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,543,521 | Supervised re-ranking for visual search | VisualRank - PageRank for images |
| US 20210019342A1 | Similar medical image search | Semantic + visual similarity embeddings |
VisualRank Concept
Applies PageRank principles to image similarity graphs:
- Images as nodes
- Visual similarity as edges
- Authority scores for image ranking
Voice Search Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 7,027,987B1 | Voice interface for a search engine | Voice query processing |
| US 8,260,809B2 | Voice-based search processing | Natural language + handwriting recognition |
| US 20110145224A1 | Speech-based incremental search | Real-time voice search |
| US 11,875,786B2 | Natural language recognition assistant | Modern voice assistant NLP |
Neural Network & AI Patents
| Patent | Title | Key Innovation |
|---|---|---|
| CA3138920A1 | Neural network for search retrieval and ranking | Deep learning for search |
| US 11,204,968B2 | Embedding layer in neural network for ranking | Embedding-based ranking |
| US 10,628,432B2 | Personalized deep models for smart suggestions | Personalized neural ranking |
| US 11,769,017B1 | Generative summaries for search results | SGE/AI Overviews foundation |
| US 20240256582A1 | Search with Generative Artificial Intelligence | Generative AI search |
SGE Patent Key Points
The "Generative Summaries" patent (US 11,769,017B1) describes:
- Selective use of LLMs for query responses
- Natural language summary generation
- Integration with traditional search results
Anchor Text Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 20060074871A1 | Incorporating anchor text into ranking | Weighted term frequency for anchors |
| US 20070143282A1 | Anchor text summarization for corroboration | Using anchor text to verify document relevance |
Content Quality Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,352,386B2 | Identifying training documents for content classifier | Training helpful content classifiers |
| US 20140280011A1 | Predicting Site Quality | Site-wide quality prediction |
Helpful Content Signals
The content classifier patents describe:
- Thin content detection
- Duplicate content identification
- User satisfaction prediction
- Site-wide quality scoring
Spam Detection Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 20240020476A1 | Determining linked spam content | SpamBrain-era link spam detection |
| US 8,224,905B2 | Spam filtration utilizing sender activity data | ML-based spam probability |
Semantic Analysis Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,312,021B2 | Generalized latent semantic analysis | LSA for document similarity |
| US 7,152,065B2 | Distributed latent semantic indexing | Scalable LSI for large datasets |
Question Answering Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 9,471,668B1 | Question-answering system | People Also Ask foundation |
| US 9,940,367B1 | Scoring candidate answer passages | Passage ranking for Q&A |
PAA Algorithm
The question-answering system:
- Analyzes initial query as question
- Identifies related questions from query logs
- Ranks questions by relevance
- Expands dynamically based on user interaction
Re-Ranking Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,126,883B2 | Re-ranking search results | Post-retrieval result reordering |
| US 20180101537A1 | Hierarchically Organized Machine Learning Models | Model hierarchy for ranking |
2023-2024 Modern Patents
| Patent | Year | Key Innovation |
|---|---|---|
| US 11,403,301B2 | 2022 | Search improvements |
| US 11,442,946B2 | 2022 | Query processing |
| US 11,663,277B2 | 2023 | Result ranking |
| US 11,693,863B1 | 2023 | Search optimization |
| US 11,694,034B2 | 2023 | Content understanding |
| US 11,706,318B2 | 2023 | Query matching |
| US 11,720,920B1 | 2023 | Result generation |
| US 11,727,046B2 | 2023 | Semantic analysis |
| US 11,734,287B2 | 2023 | Content classification |
| US 11,741,191B1 | 2023 | Search relevance |
| US 11,743,522B2 | 2023 | Query understanding |
| US 11,758,237B2 | 2023 | Entity processing |
| US 11,782,970B2 | 2023 | Result ranking |
| US 11,782,998B2 | 2023 | Content analysis |
| US 11,816,114B1 | 2023 | Search features |
| US 11,829,373B2 | 2023 | Query processing |
| US 11,836,177B2 | 2023 | Content ranking |
| US 20230394072A1 | 2023 | AI search |
| US 20230409653A1 | 2023 | Image search |
YouTube & Video Ranking Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,868,481B2 | Video recommendation based on co-occurrence | Videos watched together influence recommendations |
| US 10,387,431B2 | Video recommendation based on video titles | Title similarity for recommendations |
| US 9,106,958B2 | Video recommendation based on affect | Emotional response (smiles, engagement) influences ranking |
| US 9,503,786B2 | Video recommendation using affect | Facial expression analysis for recommendations |
| US 10,289,898B2 | Video recommendation via affect | Live-stream emotional analysis |
| US 8,589,434B2 | Recommendations based on topic clusters | YouTube topic clustering for recommendations |
| US 8,713,618B1 | Segmenting video based on timestamps | Comment timestamps identify video segments |
| US 9,552,555B1 | Content item recommendations | Deep neural network recommendations |
YouTube Ranking Algorithm Key Insights
The YouTube recommendation system uses multiple signals:
Key Ranking Factors (Patent-Based):
- Watch Time (US 9,098,511) - Primary ranking signal
- Video Co-occurrence (US 8,868,481B2) - Videos watched in same session
- Emotional Response (US 9,106,958B2) - Viewer facial expressions
- Topic Clustering (US 8,589,434B2) - Related topic grouping
- Title Similarity (US 10,387,431B2) - Semantic title matching
Social Media & Social Signals Patents
| Patent | Title | Key Innovation |
|---|---|---|
| US 8,756,239B2 | Real-time content searching in social network | Social connection strength affects ranking |
| US 9,081,823B2 | Ranking social network objects | Object-type specific ranking in social networks |
| US 8,495,058B2 | Filtering social search results | Social signals adjust search rankings |
| US 9,904,703B1 | Content based on social network interactions | Social engagement signals for search |
| US 20160246789A1 | Searching prominent users in social networks | Prominent user content prioritization |
| US 9,324,112B2 | Ranking authors in social media | Author authority in social platforms |
| US 7,788,260B2 | Ranking based on social network clicks | Social graph click frequency affects ranking |
| US 20130031106A1 | Social network query suggestions | Social-powered search suggestions |
Social Signals Impact on Search
Important Note
Google has publicly stated social signals are NOT direct ranking factors. However, these patents show Google can and does incorporate social data in various ways.
How Social Signals May Influence Search:
Patent-Documented Social Mechanisms:
- Social Graph Filtering (US 8,495,058B2) - Content from connections ranks higher for that user
- Prominent User Content (US 20160246789A1) - Influencer content may be prioritized
- Author Authority (US 9,324,112B2) - Social author scores affect visibility
- Click Patterns (US 7,788,260B2) - Social network member clicks influence ranking
Summary: Total Patent Count
| Category | Original | Additional | Total |
|---|---|---|---|
| Semantic Search | 2 | 2 | 4 |
| Phrase-Based Indexing | 15 | 0 | 15 |
| Entity & Knowledge Graph | 13 | 3 | 16 |
| Query Understanding | 8 | 4 | 12 |
| Content Quality | 5 | 2 | 7 |
| PageRank & Links | 6 | 3 | 9 |
| User Behavior | 8 | 0 | 8 |
| Modern 2023+ | 8 | 19 | 27 |
| Trust & Authority | 0 | 3 | 3 |
| Freshness/QDF | 0 | 1 | 1 |
| Crawling & Indexing | 0 | 3 | 3 |
| Local Search | 0 | 3 | 3 |
| Duplicate Detection | 0 | 4 | 4 |
| Visual/Image Search | 0 | 2 | 2 |
| Voice Search | 0 | 4 | 4 |
| Neural/AI | 0 | 5 | 5 |
| Anchor Text | 0 | 2 | 2 |
| Spam Detection | 0 | 2 | 2 |
| Semantic Analysis | 0 | 2 | 2 |
| Question Answering | 0 | 2 | 2 |
| Re-Ranking | 0 | 2 | 2 |
| YouTube & Video | 0 | 8 | 8 |
| Social Media/Signals | 0 | 8 | 8 |
| TOTAL | 65 | 84 | 149 |
Next Steps
- Return to Module 1 to start from the beginning
- View Quick Reference for all patents in one place
- Use the chat widget to ask questions about any patent