您当前的系统语言是中文，可以点击进入 https://www.ultipa.cn 获得更好的体验
Change Password
Change Email
Change Nickname
Your privacy
By clicking “Accept all cookies”, you agree Ultipa can store cookies on your device and disclose information in accordance with our Cookie Policy.
Pain Points (Case Study)
To understand the value of graph system powered search & recommendation solution, you must know the status and problem with existing/traditional search & recommendation systems, most of which share the following characteristics, it’s clear to see that the existing search or recommendation is NOT smart!
Keyword Ranking Only
Today’s web search engines can’t handle or recognize any correlation/causality between multiple keywords. They have no sense of causality or clue of intention as human beings naturally do. Therefore, the search results are singularly based on keyword ranking mechanism without much intelligence.
Sluggish Training
Collaborative filtering- based recommendation systems have to curate large volumes of data so to create pre-calculated results, and the latency can be many hours if not days. Tons of redundant training data have to be constructed which are considered waste of storage and computing power.
Bloated & Complicated Architecture
Most of today’s web-based search engine and recommendation systems are built on top of big-data frameworks, which can easily occupy hundreds of, if not thousands of, servers. Not only the data volume is bloated, the efficiency and efficacy of the system is questionable if you look at on average the throughput and ROI of each server. Most enterprises simply do NOT need to waste that many IT resources on these pointless big-data frameworks and dumb search & recommendation solutions. A more efficient and intelligent solution is desired.
Innovation
The power of graph computing and graph DBMS are best manifested when you are trying to find the correlations of different data points, or to penetrate an entity on the surface and explore deeply under the hood to understand the entire layout and structure of the “iceberg” below the surface, or to performance a high-dimensional search with context, intention, and causality.
Graph Augmented Intelligence
Graph-powered Smart Search (High-Dimensional/Causality Search)
Taking the above diagram as an example, assuming you are searching for 2 keywords (2 entities, Genghis Khan and Isaac Newton) with a graph search/recommendation system, the results illustrated is a causality path connecting the 2 parties. Clearly, existing web2.0 search engine like Google or Bing can’t handle such kind of “ridiculously smart search”.
Real-time Smart Recommendation w/ Ultipa Graph
On the other hand, the above diagram illustrates and visualizes how the graph-powered recommendation system works in an XAI (eXplainable AI) fashion:
- User A browses (favorites, adds-to-cart or purchases) Product A;
- Product A is also browsed (or purchased) by Users B, C, D...
- User B and C also browse other products like Product B.
- User C and D also browse Product D and other products.
- By aggregating and ranking these data, you have candidate products B and D.
- Now, looking into other factors like how Product B or D is related (similarity) to Product A, this can be expanded into a more sophisticated and very human-decision-making and reasoning framework, in this demo, we are using a very simplistic way to make the decision: Product D and Product A shares certain attribute (note their types can be different - so that we don't recommend another refrigerator after you have just purchased a refrigerator, which is both dumb and annoying to the customer) - in this case, Product A is a camera, Product D is a camera accessory kit, which may be something that the customer is looking for right after having his camera.
White-box & XAI
The bullet points above describe how collaborative filtering (hereafter CF) is conducted in a white- box and XAI (eXplainable AI) way. In Ultipa Graph, realizing CF is both easy and fast, it does NOT require any data training, nor does it contain any black-box components, therefore highly explainable.
Graph-powered CF is straightforward, it includes 5 steps:
- Graph data modeling: essentially building a networked dataset with users and products and how they interact with each other (user browse/favor/purchase a product!).
- Starting from a user, find all products that he acted upon (browsed, favored, or purchased).
- Find all users that have had similar actions on these products.
- Find all other products that are taken actions by users in Step 2.
- Group by and order by, and factor in other filtering or aggregation rules for products to be recommended back to THE user.
Template-based Graph Query for Collaborative Filtering
These 3 steps (Step 1-3) can be done in one amazingly simple Ultipa GQL:
Real-world search and recommendation systems tend to have a lot of bells and whistles and engineering tweaks, but the above example can show case how graph technology can be disruptive innovative, not just by performance, also but its intuitiveness. You can’t beat a system that closely imitates how human beings think and act.
Significance
Graph-based Search & Recommendation has the following advantages:Truly smart search is possible, be it multiple keywords correlation or linkage search, causality search, or some other types of search traditional search can’t handle;Real-time recommendation is possible as real-time data refreshing is made possible;Working with Knowledge Graph, such as Merchandise Knowledge Graph, the recommendation is very much human like - 100% intelligent, instead of relying on pure aggregated statistical data results;Recommendation Graph = Real-time Merchandise Graph + Customer 360-degree Graph, it offers unified all-in-one recommendation solution.
If you are going to build a search and recommendation system that’s fast and smart, with low TCO and high ROI, adopting a graph-powered solution makes every sense, welcome to the rapidly evolving world of IT - graph is on the way to become main-stream.