Cognitive Database: An Apache Spark-Based AI-Enabled Relational Database System with Rajesh Bordawekar
1 like497 views
This document presents a cognitive database system developed by Rajesh Bordawekar at IBM, utilizing Apache Spark and word embeddings to facilitate semantic queries in SQL. It outlines the design, implementation details, and various applications of the system, which allows relational databases to capture semantic similarities and perform advanced cognitive intelligence queries. A case study demonstrates the database's capabilities in handling multi-modal data, such as text and images, showcasing its integration of AI within the relational database ecosystem.
![17#AI5SAIS
#AI5SAIS
CI Semantic Clustering Query: Find all images whose similarity to user chosen images of [lion,
vulture, shark] using the attributeSimAvg UDF with similarity score greater than 0.75
SELECT X.imagename, X.classA, X.classB, X.classC, X.classD,
FROM ImageDataTable X
WHERE (X.imagename <> ’n01314663_7147.jpeg’) AND (X.imagename <> ’n01323781_13094.jpeg’) AND
(X.imagename <> ’n01314663_8531.jpeg’) AND
(attributeSimAvgUDF(’n01314663_7147.jpeg’, ’n01323781_13094.jpeg’, ’n01314663_8531.jpeg’, X.imagename) > 0.75)](https://image.slidesharecdn.com/5rajeshbordawekar-180612004257/85/Cognitive-Database-An-Apache-Spark-Based-AI-Enabled-Relational-Database-System-with-Rajesh-Bordawekar-17-320.jpg)
![19#AI5SAIS
#AI5SAIS
CI Analogy Query: Find all images whose classD satisfies the analogy query [reptile:
monitor_lizard :: aquatic_vertebrate : ?] using analogyQuery UDF having similarity score
greater than 0.5.
SELECT X.imagename, X.classA, X.classB, X.classC, X.classD
FROM ImageDataTable X
WHERE (analogyQuery(’reptile’,’monitor_lizard’,’aquatic_vertebrate’,X.classD,1) > 0.5)
X.Imagename X.classB X.classC X.classD
n02512053_1493 aquatic_vertebrate spiny_finned_fish permit, archerfish
n02512053_3292 aquatic_vertebrate spiny_finned_fish archerfish, mojarra
n02512053_602 aquatic_vertebrate spiny_finned_fish lookdown, permit](https://image.slidesharecdn.com/5rajeshbordawekar-180612004257/85/Cognitive-Database-An-Apache-Spark-Based-AI-Enabled-Relational-Database-System-with-Rajesh-Bordawekar-19-320.jpg)
Cognitive Database: An Apache Spark-Based AI-Enabled Relational Database System with Rajesh Bordawekar
- 1. Rajesh Bordawekar IBM T. J. Watson Research Center [email protected] Cognitive Database: An Apache Spark-Based AI-Enabled Relational Database System #AI5SAIS
- 2. Outline • Word Embedding Overview • Cognitive Database Design • Cognitive Intelligence (CI) Queries • Spark Implementation Details • Case Study: Image and Text Database • Summary 2#AI5SAIS #AI5SAIS
- 3. Word Embedding Overview • Unsupervised neural network based NLP approach to capture meanings of words based neighborhood context – Meaning is captured as collective contributions from words in the neighborhood • Generates semantic representation of words as low- dimensional vectors (200-300 dimensions) • Semantic similarity measured using distance metric (e.g., cosine distance) between vectors 3#AI5SAIS #AI5SAIS
- 4. Cognitive Database Key Ideas • Uses dual view of relational data: tables and meaningful text, with all relational entities mapped to text, without loss of information • Uses word-embedding approach to extract latent information from database tables • Classical Word embedding model extended to capture constraints of the relational model (e.g., primary keys) • Enables relational databases to capture and exploit semantic contextual similarities 4#AI5SAIS #AI5SAIS
- 5. Structured Data sources Relational Tables Word Embedding Model External Unstructured and Structured Data sources Cognitive Intelligence Queries in Structured Query Systems Structured Results Relational Tables Pre-trained Model Model built from data source being queried Using Embedding Models #AI5SAIS #AI5SAIS
- 6. Cognitive Database Features • Enables SQL-based information retrieval based on semantic context, rather than, data values • Unlike analytics databases, does not view database tables as feature and model repositories • Latent features exposed to users via standard SQL based Cognitive Intelligence (CI) queries • Users can invoke standard SQL queries using typed relational variables over a semantic model built over untyped strings 6#AI5SAIS #AI5SAIS
- 7. 7#AI5SAIS #AI5SAIS custID Date Merchant State Category Items Amount custA 9/16 Whole Foods NJ Fresh Produce Bananas, Apples 200 custB custC custD 10/16 10/16 9/16 Target Trader Joes Walmart Stationery Fresh Produce Stationery Bananas, Oranges Crayons, Pens, Notebooks Crayons, Folders 60 80 25 NY CT NY “custD 9/16 Walmart NY Stationery ‘Crayons, Folders’ 25” Text representation of a table row Words in the neighborhood contribute to the overall meaning of “custID” custA custC custB custD For this relational view, custA is similar to custC custB is similar to custD Words similar in meaning closer in vector space Customer Analytics Workload Meaning vector for every token
- 8. Cognitive Intelligence Queries • Semantic Similarity/Dissimilarities • Semantic Clustering • Cognitive OLAP queries • Inductive Reasoning queries • Semantic Relational Operations 8#AI5SAIS #AI5SAIS Can work with externally trained models and over multiple data types.
- 9. CI Query Example 9#AI5SAIS #AI5SAIS val result_df = spark.sql(s””” SELECT VENDOR_NAME, proximityCust_NameUDF(VENDOR_NAME, ‘$v’) AS proximityValue FROM Index_view HAVING proximityValue > 0.5 ORDER BY proximityValue DESC ”””) CI similarity Query: Find similar entities to a given entity (VENDOR_NAME) based on transaction characteristic similarities Cognitive UDF • Operates on relational variables. Can be sets or sequences • For each input variable, fetches vectors from the embedding model • Computes semantic similarity between vectors using nearest neighbor approaches
- 10. Cognitive Database Applications • Analysis over multi-modal data (Retail, Health, Insurance) • Entity similarity queries (Customer Analytics, IT Ticket Management, Time-series) • Cognitive OLAP (Finance, Insurance…) • Entity Resolution (Master Data Management) • Analysis of time-series data (IoT, Health) 10#AI5SAIS #AI5SAIS
- 11. Cognitive Databases Stages VectorDomain Learned Vectors Pre-computed External Learned Vectors UDFs External Text Sources Tokenized Relations Relational Tables Relational System Tables CI Queries Relations Cognitive ETL Vector Storage Query Execution RelationalTextDomain #AI5SAIS #AI5SAIS
- 12. Training from source database Relational Tables Data Cleaning Training Text File Word Embedding Training Word Embedding Model Create unique tokens (Python) k-means clustering (Numpy/Scipy) Create unique tokens (Python) text numerical values images Get image tags (Watson VRS) Create image features Window size Vector Dimensions Hyperparameter Tuning* #AI5SAIS #AI5SAIS
- 13. Why Spark? 2.2.0 Dataframes-based Representation Spark SQL based Cognitive Intelligence Queries (IBM Z zOS/zLinux, IBM P Linux,AIX, x86) Relational Databases CSV Files …. JSON Database Community Data Science Community Spark SQL+UDFs (Scala/Python) PySpark/Pandas APIs via Jupyter ….. GPU Acceleration Flexibility over multiple input data formats Portability across multiple platforms/OS Support for Standardized SQL Queries Usability across multiple user domains Opportunities for Acceleration #AI5SAIS #AI5SAIS
- 14. 14#AI5SAIS #AI5SAIS Cognitive Database: Spark Execution Flow SELECT X.custID, X.custName, proximityAvg(X.InvestType,Y.InvestType) FROM cust X, cust Y WHERE Y.custID=‘471’ AND proximityAvg(X.InvestType,Y.InvestType) LIMIT 5 SQL Query Similarity Computation Output TableTrained ModelInput Table Spark SQL UDF Nearest Neighbor Spark DF Spark DF Specialized Word Embedding Spark SQL Source Data Spark DF
- 15. Invoking Cognitive Database in Jupyter #AI5SAIS #AI5SAIS
- 16. 16#AI5SAIS #AI5SAIS Picture ID National Park Country Path of JPEG Image PK_01 Corbett India ./Img_Folder/Img_01.JPEG PK_05 Kruger South Africa ./Img_Folder/Img_05.JPEG PK_09 Sunderbans India ./Img_Folder/Img_09.JPEG PK_11 Serengeti Tanzania ./Img_Folder/Img_11.JPEG Picture Id Image Id National Park Country Animal Name Class Dietary Habit color PK_01 Img_01.JPEG Corbett India Elephant Mammal Herbivores Gray PK_05 Img_05.JPEG Kruger South Africa Rhinoceros Mammal Herbivores Gray PK_09 Img_09.JPEG Sunderbans India Crocodile Reptile Carnivorous Gray PK_11 Img_11.JPEG Serengeti Tanzania Lion Mammal Carnivorous Yellow Case Study: Application Database with links to images Internal Training database with features extracted from linked images The above merged data is used as an input to train the word embedding model that generates embeddings of each unique token based on the neighborhood. Each row of the database is viewed as a sentence.
- 17. 17#AI5SAIS #AI5SAIS CI Semantic Clustering Query: Find all images whose similarity to user chosen images of [lion, vulture, shark] using the attributeSimAvg UDF with similarity score greater than 0.75 SELECT X.imagename, X.classA, X.classB, X.classC, X.classD, FROM ImageDataTable X WHERE (X.imagename <> ’n01314663_7147.jpeg’) AND (X.imagename <> ’n01323781_13094.jpeg’) AND (X.imagename <> ’n01314663_8531.jpeg’) AND (attributeSimAvgUDF(’n01314663_7147.jpeg’, ’n01323781_13094.jpeg’, ’n01314663_8531.jpeg’, X.imagename) > 0.75)
- 18. 18#AI5SAIS #AI5SAIS X.Imagename X.classB X.classC X.classD n01604330_12473 bird_of_prey, mammal new_world_vulture, carnivore andean_condor, condor, sloth_bear n01316422_1684 mammal, bird_of_prey carnivore, eagle glutton_wolverine, piste_ski_run, downhill_skiing, ern, ski_slope n01324431_7056 bird_of_prey, mammal new_world_vulture, carnivore andean_condor, tayra n01604330_12473 n01316422_1684 Output n01324431_7056
- 19. 19#AI5SAIS #AI5SAIS CI Analogy Query: Find all images whose classD satisfies the analogy query [reptile: monitor_lizard :: aquatic_vertebrate : ?] using analogyQuery UDF having similarity score greater than 0.5. SELECT X.imagename, X.classA, X.classB, X.classC, X.classD FROM ImageDataTable X WHERE (analogyQuery(’reptile’,’monitor_lizard’,’aquatic_vertebrate’,X.classD,1) > 0.5) X.Imagename X.classB X.classC X.classD n02512053_1493 aquatic_vertebrate spiny_finned_fish permit, archerfish n02512053_3292 aquatic_vertebrate spiny_finned_fish archerfish, mojarra n02512053_602 aquatic_vertebrate spiny_finned_fish lookdown, permit
- 20. 20#AI5SAIS #AI5SAIS CI Query using external knowledge base: Find all images of animals whose classD similarity score to the Concept of ‘‘Hypercarnivore" of Wikipedia using proximityAvgForExtKB UDF is greater than 0.5. Exclude images that are already tagged as carnivore, herbivore, omnivore or scavenger. SELECT X.imagename,X.classA,X.classB,X.classC, X.classD FROM ImageDataTable X WHERE (proximityAvgAdvForExtKB(’CONCEPT_Hypercarnivore’, X.classD) > 0.5) ORDER BY SimScore DESC
- 21. Summary • Novel relational database system that uses word embedding approach to enable semantic queries in SQL • Spark-based implementation that loads data from a variety of sources and invokes Cognitive Intelligence queries using Spark SQL • Demonstration of the cognitive database capabilities using a multi-modal (text+image) dataset • Illustration of seamlessly integrating AI capabilities into relational database ecosystem 21#AI5SAIS #AI5SAIS
- 22. References • Bordawekar and Shmueli, Enabling Cognitive Intelligence Queries in Relational Databases using Low-dimensional Word Embeddings, arXiv:1603.07185, March 2016 • Bordawekar, Bandopadhyay, and Shmueli, Cognitive Database: A Step Towards Endowing Relational Databases with Artificial Intelligence Capabilities, arXiv:1712:07199, December 2017 22#AI5SAIS #AI5SAIS