Solution for automatic update of Chinese word segmentation full-text index in NEO4J

Posted by frog_ on Mon, 06 May 2019 13:10:03 +0200

Solution for automatic update of Chinese word segmentation full-text index in NEO4J

Failed to implement automatic updates using the NEO4J INDEX API, converting a way of thinking to solve this problem (synchronizing updates to the corresponding full-text index when updating a node or creating a new one.)

1. Sample data

Sample Data Format Reference

2. Differences between English and Chinese Full-Text Indexes

1. Create NEO4J default index

CALL apoc.index.addAllNodes('Loc', {Loc:["description","cause","year"]})
// The following retrieval was unsuccessful:
CALL apoc.index.search('Loc', 'Loc.description:Chinese~') YIELD node RETURN node
CALL apoc.index.search('Loc', 'Loc.description:Chinese*') YIELD node RETURN node
CALL apoc.index.search('Loc', 'Loc.description:test~') YIELD node RETURN node
CALL apoc.index.search('Loc', 'Loc.description:Test Chinese~') YIELD node RETURN node

2. Delete Index

CALL apoc.index.remove('Loc')

3. Create an index that supports Chinese words

CALL zdr.index.addChineseFulltextIndex('Loc', ["description","cause","year"], 'Loc') YIELD message RETURN message
// The following retrieval was successful:
CALL apoc.index.search('Loc', 'description:Chinese~') YIELD node RETURN node
CALL apoc.index.search('Loc', 'description:Chinese*') YIELD node RETURN node
CALL apoc.index.search('Loc', 'description:test~') YIELD node RETURN node
CALL apoc.index.search('Loc', 'description:Test Chinese~') YIELD node RETURN node

3. APOC has its own English full-text indexing process (indexing can be updated automatically)

1. Add Full-Text Index

CALL apoc.index.addAllNodes('Loc', {Loc:["description","cause","year"]},{autoUpdate:true})

2. New Nodes and Attributes

CREATE (n:Loc {name:'V'})  SET n.description='Testing Chinese word segmentation, the final chapter of the duplicate show was very exciting. It is said that knowledge mapping and artificial intelligence technology were applied to that movie!',n.cause='Test the English word breaker, Mobile World Congress, the world's largest gathering for the mobile industry, ' RETURN n

3. Retrieval

Indexes can be updated automatically, but they are not friendly to Chinese retrieval, such as the following tests:

// Retrieval failed:
CALL apoc.index.search('Loc', 'Loc.cause:Test English word breakers~') YIELD node RETURN node
CALL apoc.index.search('Loc', 'Loc.description:Test Chinese word segmentation~') YIELD node RETURN node
// Retrieved successfully:
CALL apoc.index.search('Loc', 'Loc.cause:Test English word breakers*') YIELD node RETURN node
CALL apoc.index.search('Loc', 'Loc.description:Test Chinese word segmentation*') YIELD node RETURN node

4. Custom Chinese word segmentation full-text index plug-in (unsuccessful automatic index update)

The addChineseFulltextAutoIndex process succeeds in creating a full-text index to add a full-text indexing process that supports Chinese, but automatic updates are not supported for updating new attributes of nodes.

1. Add Full-Text Index

CALL zdr.index.addChineseFulltextAutoIndex('IKAnalyzer',["description","cause","year"],'Loc',{autoUpdate:'true'}) YIELD message RETURN message

2. New Nodes and Attributes

CREATE (n:Loc {name:'V'})  SET n.description='Testing Chinese word segmentation, the final chapter of the duplicate show was very exciting. It is said that knowledge mapping and artificial intelligence technology were applied to that movie!',n.cause='Test the English word breaker, Mobile World Congress, the world's largest gathering for the mobile industry, ' RETURN n

3. Retrieval

After adding a full-text search, you can retrieve:

CALL zdr.index.chineseFulltextIndexSearch('IKAnalyzer', 'description:Acridyl Aminomethane Sulfonymethoxyaniline', 100) YIELD node RETURN node

Re-index before retrieving:

CALL zdr.index.chineseFulltextIndexSearch('IKAnalyzer', 'description:test~', 100) YIELD node RETURN node

V. Label Cross-search

Add ChineseFulltextAutoIndex/addChineseFulltextIndex supports multiple tags while retrieving, using the same index name when building the index.
Tag: Loc

CALL zdr.index.addChineseFulltextAutoIndex('Loc',["description","cause","name"],'Loc',{autoUpdate:'true'}) YIELD message RETURN message

Tag: LocProvince'

CALL zdr.index.addChineseFulltextAutoIndex('Loc',["description","cause","name"],'LocProvince',{autoUpdate:'true'}) YIELD message RETURN message

Retrieve node:

CALL apoc.index.search('Loc', 'name:p~') YIELD node RETURN node

6. Custom Chinese Word Segmentation Plugin (Failed to Update Indexes Independently of Nodes)

To support single-node index updates, develop the following process.(The automatic update scheme described in the third section fails, and updates to the corresponding full-text index synchronously when updating or creating a new node.)

1. Add Full-Text Index

CALL apoc.index.remove('Loc')
CALL zdr.index.addChineseFulltextIndex('Loc',["description","cause","year"],'Loc') YIELD message RETURN message

2. Add Nodes and Attributes and Update Full-Text Index

CREATE (n:Loc {name:'V'})  SET n.description='Testing Chinese word segmentation, the final chapter of the duplicate show was very exciting. It is said that knowledge mapping and artificial intelligence technology were applied to that movie!',n.cause='Test the English word breaker, Mobile World Congress, the world's largest gathering for the mobile industry, ' RETURN n

3. Add 2 new nodes or updated attributes to the index

MATCH (n) WHERE n.name='V' WITH n CALL zdr.index.addNodeChineseFulltextIndex(n, ['description']) RETURN *

4. Retrieval

CALL zdr.index.chineseFulltextIndexSearch('Loc', 'description:Test Chinese~') YIELD node RETURN node

7. Resolve Transaction Submission Timeout

If the transaction commit timeout setting is configured, Cancel when building the index.

#********************************************************************
### Neo4j transcation timeout
###******************************************************************
#dbms.transaction.timeout=180s

Use a background script to execute the indexer:

# index.sh
#!/usr/bin/env bash
nohup /neo4j-community-3.4.9/bin/neo4j-shell -file build.cql >>indexGraph.log 2>&1 &
// build.cql
CALL zdr.index.addChineseFulltextIndex('IKAnalyzer', ['description','fullname','name','lnkurl'], 'LinkedinID') YIELD message RETURN message;

All of the above references to the NEO4J custom process

Topics: Mobile shell