|
|
| Technology |
|
Searching structured content presents very different kinds of challenges compared to searching unstructured content. The current search stack of technologies to Crawl, Index, Search and Interpret that works for web pages but not for databases:
Search
When searching database, users, especially enterprise users, ask detailed questions like "Sales, Billing Balance of Delinquent Customers with Credit Score over 700", "Dept of Defense spending between 2000 and 2008 in the northeast", and expect relevant answers. SEMANTIFI takes a knowledge base approach to understand meaning of user queries and search content in building its semantic search engine.
Next-generation search engines are widely expected to be semantic search technologies driven by natural language processing (NLP) or more broadly by knowledge base (KB). NLP and KB search engines are similar in that they both seek to understand meaning of a search query but they differ in their approach.
|
NLP engines use natural language processing to understand meaning, while KB engines apply a broader suite of knowledge base with NLP being a potential part of it. It is most similar to how we as people understand the meaning of a sentence. To illustrate, while general purpose NLP engines may understand "customers who are delinquent", it takes someone from the credit industry to understand "customers who are 3 cycles delinquent" as it is industry or vertical knowledge that "1 cycle" means "one statement cycle" or "30 days".
Knowledge is such bits and pieces of information pulled together. It can be a business model. It can be terminology or rules. It can be industry/vertical specific or general/universally applicable. Search engines driven by a broad knowledge base can offer the highest-quality results across verticals compared to Keyword or NLP driven engines. about knowledge base driven search and the future of search |
| |
| Interpret |
Web search produces a plain listing of web pages. This may be fine with documents and is the norm. Will a raw listing of all databases records be ok?
If you are trying to interpret data trends, looking through detailed data is good enough when the data is limited. However it is not practical when reviewing large quantities of detailed data with thousands of records or more.
To interpret structured content, users expect meaningful summaries of data. Relevant tabular and graphical presentations can then help convert this information into insight. Semantifi achieved this with a data aggregation & reporting engine which dynamically aggregates detailed data into relevant summaries and shows automatic data visualizations as charts and tables. |
| |
| |
Crawl
Crawling, in the traditional sense of web pages, works well to harvest keywords but not to understand structure or meaning of content. While it is sufficient for simple keyword search of web pages, it is not useful to compose complex queries to search databases. SEMANTIFI's database crawler takes a machine assisted approach to help data owners configure data sources. Using this, most common web datasets in CSV or Excel can be configured in a few minutes.
Index
Google searches billions of web pages to respond in a fraction of a second and so can Yahoo, Ask, MSN and numerous other search portals. Why is the same performance not possible from databases?
It's no secret that the key to instant web search results from today's search engines in the web index, however the same technology does not work for fast results from databases. To index structured content and make real-time investigation of even multi-terabyte databases on the web, SEMANTIFI uses an equivalent of a "web index" for databases called the Answers Catalog. |
| |
|
| |
|
|
|
